qemu_capabilities.c 81.8 KB
Newer Older
1 2 3
/*
 * qemu_capabilities.c: QEMU capabilities generation
 *
4
 * Copyright (C) 2006-2013 Red Hat, Inc.
5 6 7 8 9 10 11 12 13 14 15 16 17
 * Copyright (C) 2006 Daniel P. Berrange
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with this library.  If not, see
O
Osier Yang 已提交
19
 * <http://www.gnu.org/licenses/>.
20 21 22 23 24 25 26
 *
 * Author: Daniel P. Berrange <berrange@redhat.com>
 */

#include <config.h>

#include "qemu_capabilities.h"
27
#include "viralloc.h"
28
#include "virlog.h"
29
#include "virerror.h"
E
Eric Blake 已提交
30
#include "virfile.h"
31 32
#include "virpidfile.h"
#include "virprocess.h"
33 34 35
#include "nodeinfo.h"
#include "cpu/cpu.h"
#include "domain_conf.h"
36
#include "vircommand.h"
37
#include "virbitmap.h"
38
#include "virnodesuspend.h"
39
#include "qemu_monitor.h"
40
#include "virstring.h"
41

42
#include <fcntl.h>
43 44 45
#include <sys/stat.h>
#include <unistd.h>
#include <sys/wait.h>
46
#include <stdarg.h>
47 48 49

#define VIR_FROM_THIS VIR_FROM_QEMU

50 51 52 53
/* While not public, these strings must not change. They
 * are used in domain status files which are read on
 * daemon restarts
 */
54
VIR_ENUM_IMPL(virQEMUCaps, QEMU_CAPS_LAST,
55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91
              "kqemu",  /* 0 */
              "vnc-colon",
              "no-reboot",
              "drive",
              "drive-boot",

              "name", /* 5 */
              "uuid",
              "domid",
              "vnet-hdr",
              "migrate-kvm-stdio",

              "migrate-qemu-tcp", /* 10 */
              "migrate-qemu-exec",
              "drive-cache-v2",
              "kvm",
              "drive-format",

              "vga", /* 15 */
              "0.10",
              "pci-device",
              "mem-path",
              "drive-serial",

              "xen-domid", /* 20 */
              "migrate-qemu-unix",
              "chardev",
              "enable-kvm",
              "monitor-json",

              "balloon", /* 25 */
              "device",
              "sdl",
              "smp-topology",
              "netdev",

              "rtc", /* 30 */
92
              "vhost-net",
93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125
              "rtc-td-hack",
              "no-hpet",
              "no-kvm-pit",

              "tdf", /* 35 */
              "pci-configfd",
              "nodefconfig",
              "boot-menu",
              "enable-kqemu",

              "fsdev", /* 40 */
              "nesting",
              "name-process",
              "drive-readonly",
              "smbios-type",

              "vga-qxl", /* 45 */
              "spice",
              "vga-none",
              "migrate-qemu-fd",
              "boot-index",

              "hda-duplex", /* 50 */
              "drive-aio",
              "pci-multibus",
              "pci-bootindex",
              "ccid-emulated",

              "ccid-passthru", /* 55 */
              "chardev-spicevmc",
              "device-spicevmc",
              "virtio-tx-alg",
              "device-qxl-vga",
126 127

              "pci-multifunction", /* 60 */
128
              "virtio-blk-pci.ioeventfd",
M
Michal Privoznik 已提交
129
              "sga",
130 131
              "virtio-blk-pci.event_idx",
              "virtio-net-pci.event_idx",
132 133 134 135 136 137 138 139 140 141

              "cache-directsync", /* 65 */
              "piix3-usb-uhci",
              "piix4-usb-uhci",
              "usb-ehci",
              "ich9-usb-ehci1",

              "vt82c686b-usb-uhci", /* 70 */
              "pci-ohci",
              "usb-redir",
M
Marc-André Lureau 已提交
142
              "usb-hub",
143
              "no-shutdown",
144 145

              "cache-unsafe", /* 75 */
146
              "rombar",
J
Jim Fehlig 已提交
147
              "ich9-ahci",
148
              "no-acpi",
149
              "fsdev-readonly",
150

151
              "virtio-blk-pci.scsi", /* 80 */
152
              "blk-sg-io",
O
Osier Yang 已提交
153
              "drive-copy-on-read",
154
              "cpu-host",
155
              "fsdev-writeout",
156 157

              "drive-iotune", /* 85 */
158
              "system_wakeup",
159
              "scsi-disk.channel",
160
              "scsi-block",
161
              "transaction",
162 163 164

              "block-job-sync", /* 90 */
              "block-job-async",
165
              "scsi-cd",
166
              "ide-cd",
167
              "no-user-config",
M
Marc-André Lureau 已提交
168 169

              "hda-micro", /* 95 */
170
              "dump-guest-memory",
G
Gerd Hoffmann 已提交
171
              "nec-usb-xhci",
172
              "virtio-s390",
173
              "balloon-event",
M
Marc-André Lureau 已提交
174

R
Richa Marwaha 已提交
175
              "bridge", /* 100 */
176 177
              "lsi",
              "virtio-scsi-pci",
V
Viktor Mihajlovski 已提交
178
              "blockio",
179
              "disable-s3",
R
Richa Marwaha 已提交
180

181
              "disable-s4", /* 105 */
182
              "usb-redir.filter",
183 184
              "ide-drive.wwn",
              "scsi-disk.wwn",
185
              "seccomp-sandbox",
186 187

              "reboot-timeout", /* 110 */
188
              "dump-guest-core",
189
              "seamless-migration",
190
              "block-commit",
191
              "vnc",
192 193

              "drive-mirror", /* 115 */
194 195
              "usb-redir.bootindex",
              "usb-host.bootindex",
196
              "blockdev-snapshot-sync",
197 198 199 200 201 202
              "qxl",

              "VGA", /* 120 */
              "cirrus-vga",
              "vmware-svga",
              "device-video-primary",
203
              "s390-sclp",
204 205

              "usb-serial", /* 125 */
G
Guannan Ren 已提交
206
              "usb-net",
207
              "add-fd",
208
              "nbd-server",
209 210 211
              "virtio-rng",

              "rng-random", /* 130 */
212
              "rng-egd",
O
Olivia Yin 已提交
213 214
              "virtio-ccw",
              "dtb",
215
              "megasas",
J
Ján Tomko 已提交
216 217

              "ipv6-migration", /* 135 */
218
              "machine-opt",
L
Li Zhang 已提交
219
              "machine-usb-opt",
S
Stefan Berger 已提交
220 221
              "tpm-passthrough",
              "tpm-tis",
222 223

              "nvram",  /* 140 */
H
Han Cheng 已提交
224 225 226 227 228 229
              "pci-bridge",
              "vfio-pci",
              "vfio-pci.bootindex",
              "scsi-generic",

              "scsi-generic.bootindex", /* 145 */
230
              "mem-merge",
231
              "vnc-websocket",
O
Osier Yang 已提交
232
              "drive-discard",
233
              "mlock",
234 235

              "vnc-share-policy", /* 150 */
236 237
    );

238
struct _virQEMUCaps {
239 240
    virObject object;

241 242
    bool usedQMP;

243 244 245
    char *binary;
    time_t mtime;

246
    virBitmapPtr flags;
247 248 249 250

    unsigned int version;
    unsigned int kvmVersion;

251
    virArch arch;
252 253 254 255 256 257 258

    size_t ncpuDefinitions;
    char **cpuDefinitions;

    size_t nmachineTypes;
    char **machineTypes;
    char **machineAliases;
259
    unsigned int *machineMaxCpus;
260 261
};

262
struct _virQEMUCapsCache {
263 264
    virMutex lock;
    virHashTablePtr binaries;
265
    char *libDir;
266
    char *runDir;
267 268
    uid_t runUid;
    gid_t runGid;
269 270
};

271

272 273
static virClassPtr virQEMUCapsClass;
static void virQEMUCapsDispose(void *obj);
274

275
static int virQEMUCapsOnceInit(void)
276
{
277 278 279 280
    if (!(virQEMUCapsClass = virClassNew(virClassForObject(),
                                         "virQEMUCaps",
                                         sizeof(virQEMUCaps),
                                         virQEMUCapsDispose)))
281 282 283 284 285
        return -1;

    return 0;
}

286
VIR_ONCE_GLOBAL_INIT(virQEMUCaps)
287

288
static virArch virQEMUCapsArchFromString(const char *arch)
289 290 291 292 293 294 295 296 297 298
{
    if (STREQ(arch, "i386"))
        return VIR_ARCH_I686;
    if (STREQ(arch, "arm"))
        return VIR_ARCH_ARMV7L;

    return virArchFromString(arch);
}


299
static const char *virQEMUCapsArchToString(virArch arch)
300 301 302 303 304 305 306 307 308 309
{
    if (arch == VIR_ARCH_I686)
        return "i386";
    else if (arch == VIR_ARCH_ARMV7L)
        return "arm";

    return virArchToString(arch);
}


310
static virCommandPtr
311 312
virQEMUCapsProbeCommand(const char *qemu,
                        virQEMUCapsPtr qemuCaps,
313
                        uid_t runUid, gid_t runGid)
314 315 316
{
    virCommandPtr cmd = virCommandNew(qemu);

317 318
    if (qemuCaps) {
        if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_NO_USER_CONFIG))
319
            virCommandAddArg(cmd, "-no-user-config");
320
        else if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_NODEFCONFIG))
321 322 323 324 325
            virCommandAddArg(cmd, "-nodefconfig");
    }

    virCommandAddEnvPassCommon(cmd);
    virCommandClearCaps(cmd);
326 327
    virCommandSetGID(cmd, runGid);
    virCommandSetUID(cmd, runUid);
328 329 330 331 332

    return cmd;
}


333
static void
334 335
virQEMUCapsSetDefaultMachine(virQEMUCapsPtr qemuCaps,
                             size_t defIdx)
336
{
337 338
    char *name = qemuCaps->machineTypes[defIdx];
    char *alias = qemuCaps->machineAliases[defIdx];
339
    unsigned int maxCpus = qemuCaps->machineMaxCpus[defIdx];
340 341 342 343 344 345 346

    memmove(qemuCaps->machineTypes + 1,
            qemuCaps->machineTypes,
            sizeof(qemuCaps->machineTypes[0]) * defIdx);
    memmove(qemuCaps->machineAliases + 1,
            qemuCaps->machineAliases,
            sizeof(qemuCaps->machineAliases[0]) * defIdx);
347 348 349
    memmove(qemuCaps->machineMaxCpus + 1,
            qemuCaps->machineMaxCpus,
            sizeof(qemuCaps->machineMaxCpus[0]) * defIdx);
350 351
    qemuCaps->machineTypes[0] = name;
    qemuCaps->machineAliases[0] = alias;
352
    qemuCaps->machineMaxCpus[0] = maxCpus;
353 354
}

355 356 357 358
/* Format is:
 * <machine> <desc> [(default)|(alias of <canonical>)]
 */
static int
359 360
virQEMUCapsParseMachineTypesStr(const char *output,
                                virQEMUCapsPtr qemuCaps)
361 362 363
{
    const char *p = output;
    const char *next;
364
    size_t defIdx = 0;
365 366 367

    do {
        const char *t;
368 369
        char *name;
        char *canonical = NULL;
370 371 372 373 374 375 376 377 378 379

        if ((next = strchr(p, '\n')))
            ++next;

        if (STRPREFIX(p, "Supported machines are:"))
            continue;

        if (!(t = strchr(p, ' ')) || (next && t >= next))
            continue;

380 381
        if (VIR_STRNDUP(name, p, t - p) < 0)
            return -1;
382 383

        p = t;
384
        if ((t = strstr(p, "(default)")) && (!next || t < next))
385
            defIdx = qemuCaps->nmachineTypes;
386 387 388 389 390 391

        if ((t = strstr(p, "(alias of ")) && (!next || t < next)) {
            p = t + strlen("(alias of ");
            if (!(t = strchr(p, ')')) || (next && t >= next))
                continue;

392
            if (VIR_STRNDUP(canonical, p, t - p) < 0) {
393
                VIR_FREE(name);
394
                return -1;
395 396 397
            }
        }

398
        if (VIR_REALLOC_N(qemuCaps->machineTypes, qemuCaps->nmachineTypes + 1) < 0 ||
399 400
            VIR_REALLOC_N(qemuCaps->machineAliases, qemuCaps->nmachineTypes + 1) < 0 ||
            VIR_REALLOC_N(qemuCaps->machineMaxCpus, qemuCaps->nmachineTypes + 1) < 0) {
401 402
            VIR_FREE(name);
            VIR_FREE(canonical);
403
            return -1;
404
        }
405
        qemuCaps->nmachineTypes++;
406
        if (canonical) {
407 408
            qemuCaps->machineTypes[qemuCaps->nmachineTypes-1] = canonical;
            qemuCaps->machineAliases[qemuCaps->nmachineTypes-1] = name;
409
        } else {
410 411
            qemuCaps->machineTypes[qemuCaps->nmachineTypes-1] = name;
            qemuCaps->machineAliases[qemuCaps->nmachineTypes-1] = NULL;
412
        }
413 414
        /* When parsing from command line we don't have information about maxCpus */
        qemuCaps->machineMaxCpus[qemuCaps->nmachineTypes-1] = 0;
415 416
    } while ((p = next));

417

418
    if (defIdx)
419
        virQEMUCapsSetDefaultMachine(qemuCaps, defIdx);
420 421 422 423

    return 0;
}

424
static int
425 426
virQEMUCapsProbeMachineTypes(virQEMUCapsPtr qemuCaps,
                             uid_t runUid, gid_t runGid)
427 428
{
    char *output;
429 430
    int ret = -1;
    virCommandPtr cmd;
431
    int status;
432

433 434 435 436
    /* Make sure the binary we are about to try exec'ing exists.
     * Technically we could catch the exec() failure, but that's
     * in a sub-process so it's hard to feed back a useful error.
     */
437
    if (!virFileIsExecutable(qemuCaps->binary)) {
438
        virReportSystemError(errno, _("Cannot find QEMU binary %s"),
439
                             qemuCaps->binary);
440 441 442
        return -1;
    }

443
    cmd = virQEMUCapsProbeCommand(qemuCaps->binary, qemuCaps, runUid, runGid);
444
    virCommandAddArgList(cmd, "-M", "?", NULL);
445
    virCommandSetOutputBuffer(cmd, &output);
446

447 448
    /* Ignore failure from older qemu that did not understand '-M ?'.  */
    if (virCommandRun(cmd, &status) < 0)
449 450
        goto cleanup;

451
    if (virQEMUCapsParseMachineTypesStr(output, qemuCaps) < 0)
452
        goto cleanup;
453 454 455 456

    ret = 0;

cleanup:
457 458
    VIR_FREE(output);
    virCommandFree(cmd);
459 460 461 462 463 464

    return ret;
}


typedef int
465 466
(*virQEMUCapsParseCPUModels)(const char *output,
                             virQEMUCapsPtr qemuCaps);
467 468 469 470 471 472 473

/* Format:
 *      <arch> <model>
 * qemu-0.13 encloses some model names in []:
 *      <arch> [<model>]
 */
static int
474 475
virQEMUCapsParseX86Models(const char *output,
                          virQEMUCapsPtr qemuCaps)
476 477 478
{
    const char *p = output;
    const char *next;
479
    int ret = -1;
480 481 482

    do {
        const char *t;
483
        size_t len;
484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500

        if ((next = strchr(p, '\n')))
            next++;

        if (!(t = strchr(p, ' ')) || (next && t >= next))
            continue;

        if (!STRPREFIX(p, "x86"))
            continue;

        p = t;
        while (*p == ' ')
            p++;

        if (*p == '\0' || *p == '\n')
            continue;

501
        if (VIR_EXPAND_N(qemuCaps->cpuDefinitions, qemuCaps->ncpuDefinitions, 1) < 0)
502
            goto cleanup;
503

504 505 506 507
        if (next)
            len = next - p - 1;
        else
            len = strlen(p);
508

509 510 511 512
        if (len > 2 && *p == '[' && p[len - 1] == ']') {
            p++;
            len -= 2;
        }
513

514
        if (VIR_STRNDUP(qemuCaps->cpuDefinitions[qemuCaps->ncpuDefinitions - 1], p, len) < 0)
515
            goto cleanup;
516 517
    } while ((p = next));

518
    ret = 0;
519

520 521
cleanup:
    return ret;
522 523
}

P
Prerna Saxena 已提交
524 525 526 527
/* ppc64 parser.
 * Format : PowerPC <machine> <description>
 */
static int
528 529
virQEMUCapsParsePPCModels(const char *output,
                          virQEMUCapsPtr qemuCaps)
P
Prerna Saxena 已提交
530 531 532
{
    const char *p = output;
    const char *next;
533
    int ret = -1;
P
Prerna Saxena 已提交
534 535 536

    do {
        const char *t;
537
        size_t len;
P
Prerna Saxena 已提交
538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557

        if ((next = strchr(p, '\n')))
            next++;

        if (!STRPREFIX(p, "PowerPC "))
            continue;

        /* Skip the preceding sub-string "PowerPC " */
        p += 8;

        /*Malformed string, does not obey the format 'PowerPC <model> <desc>'*/
        if (!(t = strchr(p, ' ')) || (next && t >= next))
            continue;

        if (*p == '\0')
            break;

        if (*p == '\n')
            continue;

558
        if (VIR_EXPAND_N(qemuCaps->cpuDefinitions, qemuCaps->ncpuDefinitions, 1) < 0)
559
            goto cleanup;
P
Prerna Saxena 已提交
560

561
        len = t - p - 1;
P
Prerna Saxena 已提交
562

563
        if (VIR_STRNDUP(qemuCaps->cpuDefinitions[qemuCaps->ncpuDefinitions - 1], p, len) < 0)
564
            goto cleanup;
P
Prerna Saxena 已提交
565 566
    } while ((p = next));

567
    ret = 0;
P
Prerna Saxena 已提交
568

569 570
cleanup:
    return ret;
P
Prerna Saxena 已提交
571
}
572

573
static int
574
virQEMUCapsProbeCPUModels(virQEMUCapsPtr qemuCaps, uid_t runUid, gid_t runGid)
575 576 577
{
    char *output = NULL;
    int ret = -1;
578
    virQEMUCapsParseCPUModels parse;
579
    virCommandPtr cmd;
580

581 582 583 584 585
    if (qemuCaps->arch == VIR_ARCH_I686 ||
        qemuCaps->arch == VIR_ARCH_X86_64)
        parse = virQEMUCapsParseX86Models;
    else if (qemuCaps->arch == VIR_ARCH_PPC64)
        parse = virQEMUCapsParsePPCModels;
586
    else {
587
        VIR_DEBUG("don't know how to parse %s CPU models",
588
                  virArchToString(qemuCaps->arch));
589 590 591
        return 0;
    }

592
    cmd = virQEMUCapsProbeCommand(qemuCaps->binary, qemuCaps, runUid, runGid);
593
    virCommandAddArgList(cmd, "-cpu", "?", NULL);
594
    virCommandSetOutputBuffer(cmd, &output);
595

596
    if (virCommandRun(cmd, NULL) < 0)
597 598
        goto cleanup;

599
    if (parse(output, qemuCaps) < 0)
600 601 602 603 604 605
        goto cleanup;

    ret = 0;

cleanup:
    VIR_FREE(output);
606
    virCommandFree(cmd);
607 608 609 610 611

    return ret;
}


612
static char *
613 614
virQEMUCapsFindBinaryForArch(virArch hostarch,
                             virArch guestarch)
615 616
{
    char *ret;
617
    const char *archstr = virQEMUCapsArchToString(guestarch);
618
    char *binary;
619

620
    if (virAsprintf(&binary, "qemu-system-%s", archstr) < 0)
621 622 623 624 625 626 627 628 629 630 631
        return NULL;

    ret = virFindFileInPath(binary);
    VIR_FREE(binary);
    if (ret && !virFileIsExecutable(ret))
        VIR_FREE(ret);

    if (guestarch == VIR_ARCH_I686 &&
        !ret &&
        hostarch == VIR_ARCH_X86_64) {
        ret = virFindFileInPath("qemu-system-x86_64");
632 633
        if (ret && !virFileIsExecutable(ret))
            VIR_FREE(ret);
634
    }
635

636 637 638 639 640
    if (guestarch == VIR_ARCH_I686 &&
        !ret) {
        ret = virFindFileInPath("qemu");
        if (ret && !virFileIsExecutable(ret))
            VIR_FREE(ret);
641
    }
642

643 644 645 646 647
    return ret;
}


static bool
648 649
virQEMUCapsIsValidForKVM(virArch hostarch,
                         virArch guestarch)
650
{
651
    if (hostarch == guestarch)
652
        return true;
653 654
    if (hostarch == VIR_ARCH_X86_64 &&
        guestarch == VIR_ARCH_I686)
655 656 657 658
        return true;
    return false;
}

659
static int
660 661 662 663
virQEMUCapsInitGuest(virCapsPtr caps,
                     virQEMUCapsCachePtr cache,
                     virArch hostarch,
                     virArch guestarch)
664 665
{
    virCapsGuestPtr guest;
666
    size_t i;
667 668
    bool haskvm = false;
    bool haskqemu = false;
669 670 671
    char *kvmbin = NULL;
    char *binary = NULL;
    virCapsGuestMachinePtr *machines = NULL;
672
    size_t nmachines = 0;
673 674
    virQEMUCapsPtr qemubinCaps = NULL;
    virQEMUCapsPtr kvmbinCaps = NULL;
675 676
    int ret = -1;

J
Ján Tomko 已提交
677
    /* Check for existence of base emulator, or alternate base
678 679
     * which can be used with magic cpu choice
     */
680
    binary = virQEMUCapsFindBinaryForArch(hostarch, guestarch);
681

682
    /* Ignore binary if extracting version info fails */
683
    if (binary) {
684
        if (!(qemubinCaps = virQEMUCapsCacheLookup(cache, binary))) {
685 686 687 688
            virResetLastError();
            VIR_FREE(binary);
        }
    }
689 690

    /* qemu-kvm/kvm binaries can only be used if
691 692 693 694 695
     *  - host & guest arches match
     * Or
     *  - hostarch is x86_64 and guest arch is i686
     * The latter simply needs "-cpu qemu32"
     */
696
    if (virQEMUCapsIsValidForKVM(hostarch, guestarch)) {
697 698 699
        const char *const kvmbins[] = { "/usr/libexec/qemu-kvm", /* RHEL */
                                        "qemu-kvm", /* Fedora */
                                        "kvm" }; /* Upstream .spec */
700

701 702
        for (i = 0; i < ARRAY_CARDINALITY(kvmbins); ++i) {
            kvmbin = virFindFileInPath(kvmbins[i]);
703

704 705
            if (!kvmbin)
                continue;
706

707
            if (!(kvmbinCaps = virQEMUCapsCacheLookup(cache, kvmbin))) {
708
                virResetLastError();
709 710 711
                VIR_FREE(kvmbin);
                continue;
            }
712

713 714
            if (!binary) {
                binary = kvmbin;
715
                qemubinCaps = kvmbinCaps;
716
                kvmbin = NULL;
717
                kvmbinCaps = NULL;
718
            }
719
            break;
720 721 722 723 724 725
        }
    }

    if (!binary)
        return 0;

726
    if (access("/dev/kvm", F_OK) == 0 &&
727 728
        (virQEMUCapsGet(qemubinCaps, QEMU_CAPS_KVM) ||
         virQEMUCapsGet(qemubinCaps, QEMU_CAPS_ENABLE_KVM) ||
729
         kvmbin))
730
        haskvm = true;
731 732

    if (access("/dev/kqemu", F_OK) == 0 &&
733
        virQEMUCapsGet(qemubinCaps, QEMU_CAPS_KQEMU))
734
        haskqemu = true;
J
Jiri Denemark 已提交
735

736
    if (virQEMUCapsGetMachineTypesCaps(qemubinCaps, &nmachines, &machines) < 0)
D
Daniel P. Berrange 已提交
737
        goto error;
738 739 740 741

    /* We register kvm as the base emulator too, since we can
     * just give -no-kvm to disable acceleration if required */
    if ((guest = virCapabilitiesAddGuest(caps,
D
Daniel P. Berrange 已提交
742
                                         "hvm",
743
                                         guestarch,
744 745 746 747 748 749 750 751 752 753
                                         binary,
                                         NULL,
                                         nmachines,
                                         machines)) == NULL)
        goto error;

    machines = NULL;
    nmachines = 0;

    if (caps->host.cpu &&
J
Jiri Denemark 已提交
754
        caps->host.cpu->model &&
755
        virQEMUCapsGetCPUDefinitions(qemubinCaps, NULL) > 0 &&
756 757 758
        !virCapabilitiesAddGuestFeature(guest, "cpuselection", 1, 0))
        goto error;

759
    if (virQEMUCapsGet(qemubinCaps, QEMU_CAPS_BOOTINDEX) &&
J
Jiri Denemark 已提交
760
        !virCapabilitiesAddGuestFeature(guest, "deviceboot", 1, 0))
761 762
        goto error;

D
Daniel P. Berrange 已提交
763 764 765 766 767 768 769
    if (virCapabilitiesAddGuestDomain(guest,
                                      "qemu",
                                      NULL,
                                      NULL,
                                      0,
                                      NULL) == NULL)
        goto error;
770

D
Daniel P. Berrange 已提交
771 772 773 774 775 776 777 778
    if (haskqemu &&
        virCapabilitiesAddGuestDomain(guest,
                                      "kqemu",
                                      NULL,
                                      NULL,
                                      0,
                                      NULL) == NULL)
        goto error;
779

D
Daniel P. Berrange 已提交
780 781
    if (haskvm) {
        virCapsGuestDomainPtr dom;
782

D
Daniel P. Berrange 已提交
783
        if (kvmbin &&
784
            virQEMUCapsGetMachineTypesCaps(kvmbinCaps, &nmachines, &machines) < 0)
D
Daniel P. Berrange 已提交
785
            goto error;
786

D
Daniel P. Berrange 已提交
787 788 789 790 791 792 793 794
        if ((dom = virCapabilitiesAddGuestDomain(guest,
                                                 "kvm",
                                                 kvmbin ? kvmbin : binary,
                                                 NULL,
                                                 nmachines,
                                                 machines)) == NULL) {
            goto error;
        }
795

D
Daniel P. Berrange 已提交
796 797
        machines = NULL;
        nmachines = 0;
798 799 800

    }

801 802
    if (((guestarch == VIR_ARCH_I686) ||
         (guestarch == VIR_ARCH_X86_64)) &&
803 804 805 806
        (virCapabilitiesAddGuestFeature(guest, "acpi", 1, 1) == NULL ||
         virCapabilitiesAddGuestFeature(guest, "apic", 1, 0) == NULL))
        goto error;

807
    if ((guestarch == VIR_ARCH_I686) &&
808 809 810
        (virCapabilitiesAddGuestFeature(guest, "pae", 1, 0) == NULL ||
         virCapabilitiesAddGuestFeature(guest, "nonpae", 1, 0) == NULL))
        goto error;
811 812 813 814

    ret = 0;

cleanup:
815 816
    VIR_FREE(binary);
    VIR_FREE(kvmbin);
817 818
    virObjectUnref(qemubinCaps);
    virObjectUnref(kvmbinCaps);
819 820 821 822 823 824 825 826 827 828 829

    return ret;

error:
    virCapabilitiesFreeMachines(machines, nmachines);

    goto cleanup;
}


static int
830 831
virQEMUCapsInitCPU(virCapsPtr caps,
                   virArch arch)
832 833 834 835 836 837
{
    virCPUDefPtr cpu = NULL;
    union cpuData *data = NULL;
    virNodeInfo nodeinfo;
    int ret = -1;

838
    if (VIR_ALLOC(cpu) < 0)
839 840
        goto error;

841 842
    cpu->arch = arch;

843
    if (nodeGetInfo(&nodeinfo))
844 845 846 847 848 849
        goto error;

    cpu->type = VIR_CPU_TYPE_HOST;
    cpu->sockets = nodeinfo.sockets;
    cpu->cores = nodeinfo.cores;
    cpu->threads = nodeinfo.threads;
J
Jiri Denemark 已提交
850
    caps->host.cpu = cpu;
851

852
    if (!(data = cpuNodeData(arch))
853
        || cpuDecode(cpu, data, NULL, 0, NULL) < 0)
J
Jiri Denemark 已提交
854
        goto cleanup;
855 856 857 858

    ret = 0;

cleanup:
859
    cpuDataFree(arch, data);
860 861 862 863 864 865 866 867 868

    return ret;

error:
    virCPUDefFree(cpu);
    goto cleanup;
}


869
virCapsPtr virQEMUCapsInit(virQEMUCapsCachePtr cache)
870 871
{
    virCapsPtr caps;
872
    size_t i;
T
Tal Kain 已提交
873
    virArch hostarch = virArchFromHost();
874

T
Tal Kain 已提交
875
    if ((caps = virCapabilitiesNew(hostarch,
876
                                   1, 1)) == NULL)
877
        goto error;
878 879 880 881 882 883 884

    /* Some machines have problematic NUMA toplogy causing
     * unexpected failures. We don't want to break the QEMU
     * driver in this scenario, so log errors & carry on
     */
    if (nodeCapsInitNUMA(caps) < 0) {
        virCapabilitiesFreeNUMAInfo(caps);
885
        VIR_WARN("Failed to query host NUMA topology, disabling NUMA capabilities");
886 887
    }

T
Tal Kain 已提交
888
    if (virQEMUCapsInitCPU(caps, hostarch) < 0)
889
        VIR_WARN("Failed to get host CPU");
890

891 892
    /* Add the power management features of the host */

893
    if (virNodeSuspendGetTargetMask(&caps->host.powerMgmt) < 0)
894 895
        VIR_WARN("Failed to get host power management capabilities");

896 897 898
    virCapabilitiesAddHostMigrateTransport(caps,
                                           "tcp");

899 900 901 902
    /* QEMU can support pretty much every arch that exists,
     * so just probe for them all - we gracefully fail
     * if a qemu-system-$ARCH binary can't be found
     */
903
    for (i = 0; i < VIR_ARCH_LAST; i++)
904
        if (virQEMUCapsInitGuest(caps, cache,
T
Tal Kain 已提交
905
                                 hostarch,
906
                                 i) < 0)
907
            goto error;
908 909 910

    return caps;

911
error:
912
    virObjectUnref(caps);
913 914 915 916
    return NULL;
}


917
static int
918 919
virQEMUCapsComputeCmdFlags(const char *help,
                           unsigned int version,
920
                           bool is_kvm,
921 922 923
                           unsigned int kvm_version,
                           virQEMUCapsPtr qemuCaps,
                           bool check_yajl ATTRIBUTE_UNUSED)
924 925
{
    const char *p;
R
Richa Marwaha 已提交
926
    const char *fsdev, *netdev;
927 928

    if (strstr(help, "-no-kqemu"))
929
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_KQEMU);
930
    if (strstr(help, "-enable-kqemu"))
931
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_ENABLE_KQEMU);
932
    if (strstr(help, "-no-kvm"))
933
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_KVM);
934
    if (strstr(help, "-enable-kvm"))
935
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_ENABLE_KVM);
936
    if (strstr(help, "-no-reboot"))
937
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_REBOOT);
938
    if (strstr(help, "-name")) {
939
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NAME);
940
        if (strstr(help, ",process="))
941
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_NAME_PROCESS);
942 943
    }
    if (strstr(help, "-uuid"))
944
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_UUID);
945
    if (strstr(help, "-xen-domid"))
946
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_XEN_DOMID);
947
    else if (strstr(help, "-domid"))
948
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DOMID);
949
    if (strstr(help, "-drive")) {
950 951
        const char *cache = strstr(help, "cache=");

952
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE);
953 954
        if (cache && (p = strchr(cache, ']'))) {
            if (memmem(cache, p - cache, "on|off", sizeof("on|off") - 1) == NULL)
955
                virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_V2);
956
            if (memmem(cache, p - cache, "directsync", sizeof("directsync") - 1))
957
                virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_DIRECTSYNC);
958
            if (memmem(cache, p - cache, "unsafe", sizeof("unsafe") - 1))
959
                virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_UNSAFE);
960
        }
961
        if (strstr(help, "format="))
962
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_FORMAT);
963
        if (strstr(help, "readonly="))
964
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_READONLY);
965
        if (strstr(help, "aio=threads|native"))
966
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_AIO);
O
Osier Yang 已提交
967
        if (strstr(help, "copy-on-read=on|off"))
968
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_COPY_ON_READ);
969
        if (strstr(help, "bps="))
970
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_IOTUNE);
971 972 973 974
    }
    if ((p = strstr(help, "-vga")) && !strstr(help, "-std-vga")) {
        const char *nl = strstr(p, "\n");

975
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA);
976 977

        if (strstr(p, "|qxl"))
978
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA_QXL);
979
        if ((p = strstr(p, "|none")) && p < nl)
980
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA_NONE);
981 982
    }
    if (strstr(help, "-spice"))
983
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SPICE);
984
    if (strstr(help, "-vnc"))
985
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC);
986
    if (strstr(help, "seamless-migration="))
987
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SEAMLESS_MIGRATION);
988
    if (strstr(help, "boot=on"))
989
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_BOOT);
990
    if (strstr(help, "serial=s"))
991
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_SERIAL);
992
    if (strstr(help, "-pcidevice"))
993
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCIDEVICE);
994
    if (strstr(help, "-mem-path"))
995
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MEM_PATH);
996
    if (strstr(help, "-chardev")) {
997
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV);
998
        if (strstr(help, "-chardev spicevmc"))
999
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEVMC);
1000
    }
1001
    if (strstr(help, "-balloon"))
1002
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_BALLOON);
1003
    if (strstr(help, "-device")) {
1004
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE);
1005 1006 1007 1008
        /*
         * When -device was introduced, qemu already supported drive's
         * readonly option but didn't advertise that.
         */
1009
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_READONLY);
1010 1011
    }
    if (strstr(help, "-nodefconfig"))
1012
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NODEFCONFIG);
1013
    if (strstr(help, "-no-user-config"))
1014
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_USER_CONFIG);
1015 1016
    /* The trailing ' ' is important to avoid a bogus match */
    if (strstr(help, "-rtc "))
1017
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_RTC);
1018 1019
    /* to wit */
    if (strstr(help, "-rtc-td-hack"))
1020
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_RTC_TD_HACK);
1021
    if (strstr(help, "-no-hpet"))
1022
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_HPET);
1023
    if (strstr(help, "-no-acpi"))
1024
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_ACPI);
1025
    if (strstr(help, "-no-kvm-pit-reinjection"))
1026
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_KVM_PIT);
1027
    if (strstr(help, "-tdf"))
1028
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_TDF);
1029
    if (strstr(help, "-enable-nesting"))
1030
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NESTING);
1031
    if (strstr(help, ",menu=on"))
1032
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_BOOT_MENU);
1033
    if (strstr(help, ",reboot-timeout=rb_time"))
1034
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_REBOOT_TIMEOUT);
1035
    if ((fsdev = strstr(help, "-fsdev"))) {
1036
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV);
1037
        if (strstr(fsdev, "readonly"))
1038
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV_READONLY);
1039
        if (strstr(fsdev, "writeout"))
1040
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV_WRITEOUT);
1041
    }
1042
    if (strstr(help, "-smbios type"))
1043
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SMBIOS_TYPE);
1044
    if (strstr(help, "-sandbox"))
1045
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SECCOMP_SANDBOX);
1046

R
Richa Marwaha 已提交
1047
    if ((netdev = strstr(help, "-netdev"))) {
1048 1049
        /* Disable -netdev on 0.12 since although it exists,
         * the corresponding netdev_add/remove monitor commands
1050 1051
         * do not, and we need them to be able to do hotplug.
         * But see below about RHEL build. */
R
Richa Marwaha 已提交
1052 1053
        if (version >= 13000) {
            if (strstr(netdev, "bridge"))
1054 1055
                virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV_BRIDGE);
           virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV);
R
Richa Marwaha 已提交
1056
        }
1057 1058 1059
    }

    if (strstr(help, "-sdl"))
1060
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SDL);
1061 1062 1063
    if (strstr(help, "cores=") &&
        strstr(help, "threads=") &&
        strstr(help, "sockets="))
1064
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_SMP_TOPOLOGY);
1065 1066

    if (version >= 9000)
1067
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_COLON);
1068 1069

    if (is_kvm && (version >= 10000 || kvm_version >= 74))
1070
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNET_HDR);
1071

1072
    if (strstr(help, ",vhost=")) {
1073
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VHOST_NET);
1074 1075
    }

1076 1077
    /* Do not use -no-shutdown if qemu doesn't support it or SIGTERM handling
     * is most likely buggy when used with -no-shutdown (which applies for qemu
1078
     * 0.14.* and 0.15.0)
1079
     */
1080
    if (strstr(help, "-no-shutdown") && (version < 14000 || version > 15000))
1081
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_SHUTDOWN);
1082

1083
    if (strstr(help, "dump-guest-core=on|off"))
1084
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DUMP_GUEST_CORE);
1085

O
Olivia Yin 已提交
1086 1087 1088
    if (strstr(help, "-dtb"))
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DTB);

1089 1090 1091
    if (strstr(help, "-machine"))
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MACHINE_OPT);

L
Li Zhang 已提交
1092 1093 1094 1095
     /* USB option is supported v1.3.0 onwards */
    if (qemuCaps->version >= 1003000)
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MACHINE_USB_OPT);

1096 1097 1098 1099
    /*
     * Handling of -incoming arg with varying features
     *  -incoming tcp    (kvm >= 79, qemu >= 0.10.0)
     *  -incoming exec   (kvm >= 80, qemu >= 0.10.0)
1100 1101
     *  -incoming unix   (qemu >= 0.12.0)
     *  -incoming fd     (qemu >= 0.12.0)
1102 1103 1104 1105 1106 1107 1108
     *  -incoming stdio  (all earlier kvm)
     *
     * NB, there was a pre-kvm-79 'tcp' support, but it
     * was broken, because it blocked the monitor console
     * while waiting for data, so pretend it doesn't exist
     */
    if (version >= 10000) {
1109 1110
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_TCP);
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_EXEC);
1111
        if (version >= 12000) {
1112 1113
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX);
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD);
1114
        }
1115
    } else if (kvm_version >= 79) {
1116
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_TCP);
1117
        if (kvm_version >= 80)
1118
            virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_EXEC);
1119
    } else if (kvm_version > 0) {
1120
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_KVM_STDIO);
1121 1122 1123
    }

    if (version >= 10000)
1124
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_0_10);
1125

1126
    if (version >= 11000)
1127
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VIRTIO_BLK_SG_IO);
1128

1129 1130 1131
    /* While JSON mode was available in 0.12.0, it was too
     * incomplete to contemplate using. The 0.13.0 release
     * is good enough to use, even though it lacks one or
1132 1133 1134 1135 1136
     * two features. This is also true of versions of qemu
     * built for RHEL, labeled 0.12.1, but with extra text
     * in the help output that mentions that features were
     * backported for libvirt. The benefits of JSON mode now
     * outweigh the downside.
1137
     */
1138
#if WITH_YAJL
1139
    if (version >= 13000) {
1140
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MONITOR_JSON);
1141 1142
    } else if (version >= 12000 &&
               strstr(help, "libvirt")) {
1143 1144
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MONITOR_JSON);
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV);
1145
    }
1146 1147 1148 1149 1150 1151 1152
#else
    /* Starting with qemu 0.15 and newer, upstream qemu no longer
     * promises to keep the human interface stable, but requests that
     * we use QMP (the JSON interface) for everything.  If the user
     * forgot to include YAJL libraries when building their own
     * libvirt but is targetting a newer qemu, we are better off
     * telling them to recompile (the spec file includes the
1153
     * dependency, so distros won't hit this).  This check is
1154
     * also in m4/virt-yajl.m4 (see $with_yajl).  */
1155 1156 1157
    if (version >= 15000 ||
        (version >= 12000 && strstr(help, "libvirt"))) {
        if (check_yajl) {
1158 1159 1160
            virReportError(VIR_ERR_CONFIG_UNSUPPORTED, "%s",
                           _("this qemu binary requires libvirt to be "
                             "compiled with yajl"));
1161 1162
            return -1;
        }
1163
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV);
1164
    }
E
Eric Blake 已提交
1165
#endif
1166 1167

    if (version >= 13000)
1168
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCI_MULTIFUNCTION);
1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180

    /* Although very new versions of qemu advertise the presence of
     * the rombar option in the output of "qemu -device pci-assign,?",
     * this advertisement was added to the code long after the option
     * itself. According to qemu developers, though, rombar is
     * available in all qemu binaries from release 0.12 onward.
     * Setting the capability this way makes it available in more
     * cases where it might be needed, and shouldn't cause any false
     * positives (in the case that it did, qemu would produce an error
     * log and refuse to start, so it would be immediately obvious).
     */
    if (version >= 12000)
1181
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCI_ROMBAR);
1182 1183

    if (version >= 11000)
1184
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_CPU_HOST);
1185

1186
    if (version >= 1001000) {
J
Ján Tomko 已提交
1187
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
1188 1189
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_SHARE_POLICY);
    }
J
Ján Tomko 已提交
1190

1191
    if (version >= 1002000)
1192
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE_VIDEO_PRIMARY);
1193
    return 0;
1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219
}

/* We parse the output of 'qemu -help' to get the QEMU
 * version number. The first bit is easy, just parse
 * 'QEMU PC emulator version x.y.z'
 * or
 * 'QEMU emulator version x.y.z'.
 *
 * With qemu-kvm, however, that is followed by a string
 * in parenthesis as follows:
 *  - qemu-kvm-x.y.z in stable releases
 *  - kvm-XX for kvm versions up to kvm-85
 *  - qemu-kvm-devel-XX for kvm version kvm-86 and later
 *
 * For qemu-kvm versions before 0.10.z, we need to detect
 * the KVM version number for some features. With 0.10.z
 * and later, we just need the QEMU version number and
 * whether it is KVM QEMU or mainline QEMU.
 */
#define QEMU_VERSION_STR_1  "QEMU emulator version"
#define QEMU_VERSION_STR_2  "QEMU PC emulator version"
#define QEMU_KVM_VER_PREFIX "(qemu-kvm-"
#define KVM_VER_PREFIX      "(kvm-"

#define SKIP_BLANKS(p) do { while ((*(p) == ' ') || (*(p) == '\t')) (p)++; } while (0)

1220 1221 1222 1223
int virQEMUCapsParseHelpStr(const char *qemu,
                            const char *help,
                            virQEMUCapsPtr qemuCaps,
                            unsigned int *version,
1224
                            bool *is_kvm,
1225 1226
                            unsigned int *kvm_version,
                            bool check_yajl)
1227 1228 1229
{
    unsigned major, minor, micro;
    const char *p = help;
1230
    char *strflags;
1231

1232 1233
    *version = *kvm_version = 0;
    *is_kvm = false;
1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250

    if (STRPREFIX(p, QEMU_VERSION_STR_1))
        p += strlen(QEMU_VERSION_STR_1);
    else if (STRPREFIX(p, QEMU_VERSION_STR_2))
        p += strlen(QEMU_VERSION_STR_2);
    else
        goto fail;

    SKIP_BLANKS(p);

    major = virParseNumber(&p);
    if (major == -1 || *p != '.')
        goto fail;

    ++p;

    minor = virParseNumber(&p);
J
Jiri Denemark 已提交
1251
    if (minor == -1)
1252 1253
        goto fail;

J
Jiri Denemark 已提交
1254 1255 1256 1257 1258 1259 1260 1261
    if (*p != '.') {
        micro = 0;
    } else {
        ++p;
        micro = virParseNumber(&p);
        if (micro == -1)
            goto fail;
    }
1262 1263 1264 1265

    SKIP_BLANKS(p);

    if (STRPREFIX(p, QEMU_KVM_VER_PREFIX)) {
1266
        *is_kvm = true;
1267 1268 1269 1270
        p += strlen(QEMU_KVM_VER_PREFIX);
    } else if (STRPREFIX(p, KVM_VER_PREFIX)) {
        int ret;

1271
        *is_kvm = true;
1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282
        p += strlen(KVM_VER_PREFIX);

        ret = virParseNumber(&p);
        if (ret == -1)
            goto fail;

        *kvm_version = ret;
    }

    *version = (major * 1000 * 1000) + (minor * 1000) + micro;

1283 1284
    if (virQEMUCapsComputeCmdFlags(help, *version, *is_kvm, *kvm_version,
                                   qemuCaps, check_yajl) < 0)
1285
        goto cleanup;
1286

1287
    strflags = virBitmapString(qemuCaps->flags);
1288 1289 1290
    VIR_DEBUG("Version %u.%u.%u, cooked version %u, flags %s",
              major, minor, micro, *version, NULLSTR(strflags));
    VIR_FREE(strflags);
1291 1292 1293 1294 1295 1296 1297 1298 1299 1300

    if (*kvm_version)
        VIR_DEBUG("KVM version %d detected", *kvm_version);
    else if (*is_kvm)
        VIR_DEBUG("qemu-kvm version %u.%u.%u detected", major, minor, micro);

    return 0;

fail:
    p = strchr(help, '\n');
1301 1302
    if (!p)
        p = strchr(help, '\0');
1303

1304 1305 1306
    virReportError(VIR_ERR_INTERNAL_ERROR,
                   _("cannot parse %s version number in '%.*s'"),
                   qemu, (int) (p - help), help);
1307

1308
cleanup:
1309 1310 1311
    return -1;
}

1312

1313
struct virQEMUCapsStringFlags {
1314 1315 1316 1317 1318
    const char *value;
    int flag;
};


1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334
struct virQEMUCapsStringFlags virQEMUCapsCommands[] = {
    { "system_wakeup", QEMU_CAPS_WAKEUP },
    { "transaction", QEMU_CAPS_TRANSACTION },
    { "block_job_cancel", QEMU_CAPS_BLOCKJOB_SYNC },
    { "block-job-cancel", QEMU_CAPS_BLOCKJOB_ASYNC },
    { "dump-guest-memory", QEMU_CAPS_DUMP_GUEST_MEMORY },
    { "query-spice", QEMU_CAPS_SPICE },
    { "query-kvm", QEMU_CAPS_KVM },
    { "block-commit", QEMU_CAPS_BLOCK_COMMIT },
    { "query-vnc", QEMU_CAPS_VNC },
    { "drive-mirror", QEMU_CAPS_DRIVE_MIRROR },
    { "blockdev-snapshot-sync", QEMU_CAPS_DISK_SNAPSHOT },
    { "add-fd", QEMU_CAPS_ADD_FD },
    { "nbd-server-start", QEMU_CAPS_NBD_SERVER },
};

1335 1336 1337 1338 1339
struct virQEMUCapsStringFlags virQEMUCapsEvents[] = {
    { "BALLOON_CHANGE", QEMU_CAPS_BALLOON_EVENT },
    { "SPICE_MIGRATE_COMPLETED", QEMU_CAPS_SEAMLESS_MIGRATION },
};

1340
struct virQEMUCapsStringFlags virQEMUCapsObjectTypes[] = {
1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355
    { "hda-duplex", QEMU_CAPS_HDA_DUPLEX },
    { "hda-micro", QEMU_CAPS_HDA_MICRO },
    { "ccid-card-emulated", QEMU_CAPS_CCID_EMULATED },
    { "ccid-card-passthru", QEMU_CAPS_CCID_PASSTHRU },
    { "piix3-usb-uhci", QEMU_CAPS_PIIX3_USB_UHCI },
    { "piix4-usb-uhci", QEMU_CAPS_PIIX4_USB_UHCI },
    { "usb-ehci", QEMU_CAPS_USB_EHCI },
    { "ich9-usb-ehci1", QEMU_CAPS_ICH9_USB_EHCI1 },
    { "vt82c686b-usb-uhci", QEMU_CAPS_VT82C686B_USB_UHCI },
    { "pci-ohci", QEMU_CAPS_PCI_OHCI },
    { "nec-usb-xhci", QEMU_CAPS_NEC_USB_XHCI },
    { "usb-redir", QEMU_CAPS_USB_REDIR },
    { "usb-hub", QEMU_CAPS_USB_HUB },
    { "ich9-ahci", QEMU_CAPS_ICH9_AHCI },
    { "virtio-blk-s390", QEMU_CAPS_VIRTIO_S390 },
1356
    { "virtio-blk-ccw", QEMU_CAPS_VIRTIO_CCW },
1357
    { "sclpconsole", QEMU_CAPS_SCLP_S390 },
1358
    { "lsi53c895a", QEMU_CAPS_SCSI_LSI },
1359
    { "virtio-scsi-pci", QEMU_CAPS_VIRTIO_SCSI },
1360 1361
    { "virtio-scsi-s390", QEMU_CAPS_VIRTIO_SCSI },
    { "virtio-scsi-ccw", QEMU_CAPS_VIRTIO_SCSI },
1362
    { "megasas", QEMU_CAPS_SCSI_MEGASAS },
1363 1364
    { "spicevmc", QEMU_CAPS_DEVICE_SPICEVMC },
    { "qxl-vga", QEMU_CAPS_DEVICE_QXL_VGA },
1365
    { "qxl", QEMU_CAPS_DEVICE_QXL },
1366 1367 1368 1369
    { "sga", QEMU_CAPS_SGA },
    { "scsi-block", QEMU_CAPS_SCSI_BLOCK },
    { "scsi-cd", QEMU_CAPS_SCSI_CD },
    { "ide-cd", QEMU_CAPS_IDE_CD },
1370 1371 1372
    { "VGA", QEMU_CAPS_DEVICE_VGA },
    { "cirrus-vga", QEMU_CAPS_DEVICE_CIRRUS_VGA },
    { "vmware-svga", QEMU_CAPS_DEVICE_VMWARE_SVGA },
H
Han Cheng 已提交
1373 1374
    { "usb-serial", QEMU_CAPS_DEVICE_USB_SERIAL },
    { "usb-net", QEMU_CAPS_DEVICE_USB_NET },
1375
    { "virtio-rng-pci", QEMU_CAPS_DEVICE_VIRTIO_RNG },
1376 1377
    { "virtio-rng-s390", QEMU_CAPS_DEVICE_VIRTIO_RNG },
    { "virtio-rng-ccw", QEMU_CAPS_DEVICE_VIRTIO_RNG },
1378
    { "rng-random", QEMU_CAPS_OBJECT_RNG_RANDOM },
1379
    { "rng-egd", QEMU_CAPS_OBJECT_RNG_EGD },
1380
    { "spapr-nvram", QEMU_CAPS_DEVICE_NVRAM },
1381
    { "pci-bridge", QEMU_CAPS_DEVICE_PCI_BRIDGE },
1382
    { "vfio-pci", QEMU_CAPS_DEVICE_VFIO_PCI },
H
Han Cheng 已提交
1383
    { "scsi-generic", QEMU_CAPS_DEVICE_SCSI_GENERIC },
1384 1385
};

1386
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsVirtioBlk[] = {
1387 1388 1389 1390 1391 1392 1393 1394
    { "multifunction", QEMU_CAPS_PCI_MULTIFUNCTION },
    { "bootindex", QEMU_CAPS_BOOTINDEX },
    { "ioeventfd", QEMU_CAPS_VIRTIO_IOEVENTFD },
    { "event_idx", QEMU_CAPS_VIRTIO_BLK_EVENT_IDX },
    { "scsi", QEMU_CAPS_VIRTIO_BLK_SCSI },
    { "logical_block_size", QEMU_CAPS_BLOCKIO },
};

1395
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsVirtioNet[] = {
1396 1397 1398 1399
    { "tx", QEMU_CAPS_VIRTIO_TX_ALG },
    { "event_idx", QEMU_CAPS_VIRTIO_NET_EVENT_IDX },
};

1400
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsPciAssign[] = {
1401
    { "rombar", QEMU_CAPS_PCI_ROMBAR },
1402 1403 1404 1405
    { "configfd", QEMU_CAPS_PCI_CONFIGFD },
    { "bootindex", QEMU_CAPS_PCI_BOOTINDEX },
};

1406 1407 1408 1409
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsVfioPci[] = {
    { "bootindex", QEMU_CAPS_VFIO_PCI_BOOTINDEX },
};

1410
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsScsiDisk[] = {
1411 1412 1413 1414
    { "channel", QEMU_CAPS_SCSI_DISK_CHANNEL },
    { "wwn", QEMU_CAPS_SCSI_DISK_WWN },
};

1415
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsIDEDrive[] = {
1416 1417 1418
    { "wwn", QEMU_CAPS_IDE_DRIVE_WWN },
};

1419
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsPixx4PM[] = {
1420 1421 1422 1423
    { "disable_s3", QEMU_CAPS_DISABLE_S3 },
    { "disable_s4", QEMU_CAPS_DISABLE_S4 },
};

1424
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsUsbRedir[] = {
1425
    { "filter", QEMU_CAPS_USB_REDIR_FILTER },
1426 1427 1428
    { "bootindex", QEMU_CAPS_USB_REDIR_BOOTINDEX },
};

1429
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsUsbHost[] = {
1430
    { "bootindex", QEMU_CAPS_USB_HOST_BOOTINDEX },
1431 1432
};

H
Han Cheng 已提交
1433 1434 1435 1436
static struct virQEMUCapsStringFlags virQEMUCapsObjectPropsScsiGeneric[] = {
    { "bootindex", QEMU_CAPS_DEVICE_SCSI_GENERIC_BOOTINDEX },
};

1437
struct virQEMUCapsObjectTypeProps {
1438
    const char *type;
1439
    struct virQEMUCapsStringFlags *props;
1440 1441 1442
    size_t nprops;
};

1443 1444 1445 1446
static struct virQEMUCapsObjectTypeProps virQEMUCapsObjectProps[] = {
    { "virtio-blk-pci", virQEMUCapsObjectPropsVirtioBlk,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioBlk) },
    { "virtio-net-pci", virQEMUCapsObjectPropsVirtioNet,
1447 1448 1449 1450
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioNet) },
    { "virtio-blk-ccw", virQEMUCapsObjectPropsVirtioBlk,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioBlk) },
    { "virtio-net-ccw", virQEMUCapsObjectPropsVirtioNet,
1451 1452 1453 1454 1455 1456 1457 1458 1459
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioNet) },
    { "virtio-blk-s390", virQEMUCapsObjectPropsVirtioBlk,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioBlk) },
    { "virtio-net-s390", virQEMUCapsObjectPropsVirtioNet,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVirtioNet) },
    { "pci-assign", virQEMUCapsObjectPropsPciAssign,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsPciAssign) },
    { "kvm-pci-assign", virQEMUCapsObjectPropsPciAssign,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsPciAssign) },
1460 1461
    { "vfio-pci", virQEMUCapsObjectPropsVfioPci,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsVfioPci) },
1462 1463 1464 1465 1466 1467 1468 1469 1470 1471
    { "scsi-disk", virQEMUCapsObjectPropsScsiDisk,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsScsiDisk) },
    { "ide-drive", virQEMUCapsObjectPropsIDEDrive,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsIDEDrive) },
    { "PIIX4_PM", virQEMUCapsObjectPropsPixx4PM,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsPixx4PM) },
    { "usb-redir", virQEMUCapsObjectPropsUsbRedir,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsUsbRedir) },
    { "usb-host", virQEMUCapsObjectPropsUsbHost,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsUsbHost) },
H
Han Cheng 已提交
1472 1473
    { "scsi-generic", virQEMUCapsObjectPropsScsiGeneric,
      ARRAY_CARDINALITY(virQEMUCapsObjectPropsScsiGeneric) },
1474 1475 1476 1477
};


static void
1478 1479 1480 1481 1482
virQEMUCapsProcessStringFlags(virQEMUCapsPtr qemuCaps,
                              size_t nflags,
                              struct virQEMUCapsStringFlags *flags,
                              size_t nvalues,
                              char *const*values)
1483 1484
{
    size_t i, j;
1485 1486
    for (i = 0; i < nflags; i++) {
        for (j = 0; j < nvalues; j++) {
1487
            if (STREQ(values[j], flags[i].value)) {
1488
                virQEMUCapsSet(qemuCaps, flags[i].flag);
1489 1490 1491 1492 1493 1494 1495 1496
                break;
            }
        }
    }
}


static void
1497 1498
virQEMUCapsFreeStringList(size_t len,
                          char **values)
1499 1500
{
    size_t i;
1501
    for (i = 0; i < len; i++)
1502 1503 1504 1505 1506 1507 1508 1509
        VIR_FREE(values[i]);
    VIR_FREE(values);
}


#define OBJECT_TYPE_PREFIX "name \""

static int
1510 1511
virQEMUCapsParseDeviceStrObjectTypes(const char *str,
                                     char ***types)
1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529
{
    const char *tmp = str;
    int ret = -1;
    size_t ntypelist = 0;
    char **typelist = NULL;

    *types = NULL;

    while ((tmp = strstr(tmp, OBJECT_TYPE_PREFIX))) {
        char *end;
        tmp += strlen(OBJECT_TYPE_PREFIX);
        end = strstr(tmp, "\"");
        if (!end) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Malformed QEMU device list string, missing quote"));
            goto cleanup;
        }

1530
        if (VIR_EXPAND_N(typelist, ntypelist, 1) < 0)
1531
            goto cleanup;
1532
        if (VIR_STRNDUP(typelist[ntypelist - 1], tmp, end-tmp) < 0)
1533 1534 1535 1536 1537 1538 1539 1540
            goto cleanup;
    }

    *types = typelist;
    ret = ntypelist;

cleanup:
    if (ret < 0)
1541
        virQEMUCapsFreeStringList(ntypelist, typelist);
1542 1543 1544 1545 1546
    return ret;
}


static int
1547 1548 1549
virQEMUCapsParseDeviceStrObjectProps(const char *str,
                                     const char *type,
                                     char ***props)
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582
{
    const char *tmp = str;
    int ret = -1;
    size_t nproplist = 0;
    char **proplist = NULL;

    VIR_DEBUG("Extract type %s", type);
    *props = NULL;

    while ((tmp = strchr(tmp, '\n'))) {
        char *end;
        tmp += 1;

        if (*tmp == '\0')
            break;

        if (STRPREFIX(tmp, OBJECT_TYPE_PREFIX))
            continue;

        if (!STRPREFIX(tmp, type))
            continue;

        tmp += strlen(type);
        if (*tmp != '.')
            continue;
        tmp++;

        end = strstr(tmp, "=");
        if (!end) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("Malformed QEMU device list string, missing '='"));
            goto cleanup;
        }
1583
        if (VIR_EXPAND_N(proplist, nproplist, 1) < 0)
1584
            goto cleanup;
1585
        if (VIR_STRNDUP(proplist[nproplist - 1], tmp, end-tmp) < 0)
1586 1587 1588 1589 1590 1591 1592 1593
            goto cleanup;
    }

    *props = proplist;
    ret = nproplist;

cleanup:
    if (ret < 0)
1594
        virQEMUCapsFreeStringList(nproplist, proplist);
1595 1596 1597 1598 1599
    return ret;
}


int
1600
virQEMUCapsParseDeviceStr(virQEMUCapsPtr qemuCaps, const char *str)
1601 1602 1603 1604 1605
{
    int nvalues;
    char **values;
    size_t i;

1606
    if ((nvalues = virQEMUCapsParseDeviceStrObjectTypes(str, &values)) < 0)
1607
        return -1;
1608 1609 1610 1611 1612 1613
    virQEMUCapsProcessStringFlags(qemuCaps,
                                  ARRAY_CARDINALITY(virQEMUCapsObjectTypes),
                                  virQEMUCapsObjectTypes,
                                  nvalues, values);
    virQEMUCapsFreeStringList(nvalues, values);

1614
    for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsObjectProps); i++) {
1615 1616 1617 1618
        const char *type = virQEMUCapsObjectProps[i].type;
        if ((nvalues = virQEMUCapsParseDeviceStrObjectProps(str,
                                                            type,
                                                            &values)) < 0)
1619
            return -1;
1620 1621 1622 1623 1624
        virQEMUCapsProcessStringFlags(qemuCaps,
                                      virQEMUCapsObjectProps[i].nprops,
                                      virQEMUCapsObjectProps[i].props,
                                      nvalues, values);
        virQEMUCapsFreeStringList(nvalues, values);
1625 1626 1627
    }

    /* Prefer -chardev spicevmc (detected earlier) over -device spicevmc */
1628 1629
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEVMC))
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_DEVICE_SPICEVMC);
1630 1631 1632 1633 1634

    return 0;
}


E
Eric Blake 已提交
1635
static int
1636 1637
virQEMUCapsExtractDeviceStr(const char *qemu,
                            virQEMUCapsPtr qemuCaps,
1638
                            uid_t runUid, gid_t runGid)
1639
{
E
Eric Blake 已提交
1640
    char *output = NULL;
1641
    virCommandPtr cmd;
E
Eric Blake 已提交
1642
    int ret = -1;
1643

E
Eric Blake 已提交
1644 1645
    /* Cram together all device-related queries into one invocation;
     * the output format makes it possible to distinguish what we
1646 1647
     * need.  With qemu 0.13.0 and later, unrecognized '-device
     * bogus,?' cause an error in isolation, but are silently ignored
1648
     * in combination with '-device ?'.  Upstream qemu 0.12.x doesn't
1649 1650
     * understand '-device name,?', and always exits with status 1 for
     * the simpler '-device ?', so this function is really only useful
1651
     * if -help includes "device driver,?".  */
1652
    cmd = virQEMUCapsProbeCommand(qemu, qemuCaps, runUid, runGid);
1653 1654 1655 1656 1657 1658
    virCommandAddArgList(cmd,
                         "-device", "?",
                         "-device", "pci-assign,?",
                         "-device", "virtio-blk-pci,?",
                         "-device", "virtio-net-pci,?",
                         "-device", "scsi-disk,?",
1659
                         "-device", "PIIX4_PM,?",
1660
                         "-device", "usb-redir,?",
1661
                         "-device", "ide-drive,?",
1662
                         "-device", "usb-host,?",
H
Han Cheng 已提交
1663
                         "-device", "scsi-generic,?",
1664
                         NULL);
1665
    /* qemu -help goes to stdout, but qemu -device ? goes to stderr.  */
E
Eric Blake 已提交
1666
    virCommandSetErrorBuffer(cmd, &output);
1667

1668
    if (virCommandRun(cmd, NULL) < 0)
1669 1670
        goto cleanup;

1671
    ret = virQEMUCapsParseDeviceStr(qemuCaps, output);
1672 1673

cleanup:
E
Eric Blake 已提交
1674
    VIR_FREE(output);
1675
    virCommandFree(cmd);
E
Eric Blake 已提交
1676 1677 1678
    return ret;
}

1679

1680 1681 1682
int virQEMUCapsGetDefaultVersion(virCapsPtr caps,
                                 virQEMUCapsCachePtr capsCache,
                                 unsigned int *version)
1683 1684
{
    const char *binary;
1685
    virQEMUCapsPtr qemucaps;
T
Tal Kain 已提交
1686
    virArch hostarch;
1687 1688 1689 1690

    if (*version > 0)
        return 0;

T
Tal Kain 已提交
1691
    hostarch = virArchFromHost();
1692 1693
    if ((binary = virCapabilitiesDefaultGuestEmulator(caps,
                                                      "hvm",
T
Tal Kain 已提交
1694
                                                      hostarch,
1695
                                                      "qemu")) == NULL) {
1696
        virReportError(VIR_ERR_INTERNAL_ERROR,
1697
                       _("Cannot find suitable emulator for %s"),
T
Tal Kain 已提交
1698
                       virArchToString(hostarch));
1699 1700 1701
        return -1;
    }

1702
    if (!(qemucaps = virQEMUCapsCacheLookup(capsCache, binary)))
1703 1704
        return -1;

1705
    *version = virQEMUCapsGetVersion(qemucaps);
1706
    virObjectUnref(qemucaps);
1707 1708
    return 0;
}
1709 1710


1711 1712


1713 1714
virQEMUCapsPtr
virQEMUCapsNew(void)
1715
{
1716
    virQEMUCapsPtr qemuCaps;
1717

1718
    if (virQEMUCapsInitialize() < 0)
1719 1720
        return NULL;

1721
    if (!(qemuCaps = virObjectNew(virQEMUCapsClass)))
1722 1723
        return NULL;

1724
    if (!(qemuCaps->flags = virBitmapNew(QEMU_CAPS_LAST)))
1725
        goto error;
1726

1727
    return qemuCaps;
1728

1729
error:
1730
    virObjectUnref(qemuCaps);
1731
    return NULL;
1732 1733 1734
}


1735
virQEMUCapsPtr virQEMUCapsNewCopy(virQEMUCapsPtr qemuCaps)
1736
{
1737
    virQEMUCapsPtr ret = virQEMUCapsNew();
1738 1739 1740 1741 1742
    size_t i;

    if (!ret)
        return NULL;

1743
    virBitmapCopy(ret->flags, qemuCaps->flags);
1744

1745 1746 1747 1748
    ret->usedQMP = qemuCaps->usedQMP;
    ret->version = qemuCaps->version;
    ret->kvmVersion = qemuCaps->kvmVersion;
    ret->arch = qemuCaps->arch;
1749

1750
    if (VIR_ALLOC_N(ret->cpuDefinitions, qemuCaps->ncpuDefinitions) < 0)
1751
        goto error;
1752
    ret->ncpuDefinitions = qemuCaps->ncpuDefinitions;
1753
    for (i = 0; i < qemuCaps->ncpuDefinitions; i++) {
1754 1755
        if (VIR_STRDUP(ret->cpuDefinitions[i], qemuCaps->cpuDefinitions[i]) < 0)
            goto error;
1756 1757
    }

1758
    if (VIR_ALLOC_N(ret->machineTypes, qemuCaps->nmachineTypes) < 0)
1759
        goto error;
1760
    if (VIR_ALLOC_N(ret->machineAliases, qemuCaps->nmachineTypes) < 0)
1761
        goto error;
1762
    if (VIR_ALLOC_N(ret->machineMaxCpus, qemuCaps->nmachineTypes) < 0)
1763
        goto error;
1764
    ret->nmachineTypes = qemuCaps->nmachineTypes;
1765
    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
1766 1767 1768
        if (VIR_STRDUP(ret->machineTypes[i], qemuCaps->machineTypes[i]) < 0 ||
            VIR_STRDUP(ret->machineAliases[i], qemuCaps->machineAliases[i]) < 0)
            goto error;
1769
        ret->machineMaxCpus[i] = qemuCaps->machineMaxCpus[i];
1770 1771 1772 1773
    }

    return ret;

1774
error:
1775 1776 1777 1778 1779
    virObjectUnref(ret);
    return NULL;
}


1780
void virQEMUCapsDispose(void *obj)
1781
{
1782
    virQEMUCapsPtr qemuCaps = obj;
1783 1784
    size_t i;

1785
    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
1786 1787
        VIR_FREE(qemuCaps->machineTypes[i]);
        VIR_FREE(qemuCaps->machineAliases[i]);
1788
    }
1789 1790
    VIR_FREE(qemuCaps->machineTypes);
    VIR_FREE(qemuCaps->machineAliases);
1791
    VIR_FREE(qemuCaps->machineMaxCpus);
1792

1793
    for (i = 0; i < qemuCaps->ncpuDefinitions; i++) {
1794
        VIR_FREE(qemuCaps->cpuDefinitions[i]);
1795
    }
1796
    VIR_FREE(qemuCaps->cpuDefinitions);
1797

1798
    virBitmapFree(qemuCaps->flags);
1799

1800
    VIR_FREE(qemuCaps->binary);
1801 1802
}

1803
void
1804 1805
virQEMUCapsSet(virQEMUCapsPtr qemuCaps,
               enum virQEMUCapsFlags flag)
1806
{
1807
    ignore_value(virBitmapSetBit(qemuCaps->flags, flag));
1808 1809 1810 1811
}


void
1812
virQEMUCapsSetList(virQEMUCapsPtr qemuCaps, ...)
1813 1814 1815 1816
{
    va_list list;
    int flag;

1817
    va_start(list, qemuCaps);
1818
    while ((flag = va_arg(list, int)) < QEMU_CAPS_LAST)
1819
        ignore_value(virBitmapSetBit(qemuCaps->flags, flag));
1820
    va_end(list);
1821 1822 1823 1824
}


void
1825 1826
virQEMUCapsClear(virQEMUCapsPtr qemuCaps,
                 enum virQEMUCapsFlags flag)
1827
{
1828
    ignore_value(virBitmapClearBit(qemuCaps->flags, flag));
1829 1830 1831
}


1832
char *virQEMUCapsFlagsString(virQEMUCapsPtr qemuCaps)
1833
{
1834
    return virBitmapString(qemuCaps->flags);
1835 1836 1837 1838
}


bool
1839 1840
virQEMUCapsGet(virQEMUCapsPtr qemuCaps,
               enum virQEMUCapsFlags flag)
1841
{
1842 1843
    bool b;

1844
    if (!qemuCaps || virBitmapGetBit(qemuCaps->flags, flag, &b) < 0)
1845 1846 1847
        return false;
    else
        return b;
1848
}
1849 1850


1851
const char *virQEMUCapsGetBinary(virQEMUCapsPtr qemuCaps)
1852
{
1853
    return qemuCaps->binary;
1854 1855
}

1856
virArch virQEMUCapsGetArch(virQEMUCapsPtr qemuCaps)
1857
{
1858
    return qemuCaps->arch;
1859 1860 1861
}


1862
unsigned int virQEMUCapsGetVersion(virQEMUCapsPtr qemuCaps)
1863
{
1864
    return qemuCaps->version;
1865 1866 1867
}


1868
unsigned int virQEMUCapsGetKVMVersion(virQEMUCapsPtr qemuCaps)
1869
{
1870
    return qemuCaps->kvmVersion;
1871 1872 1873
}


1874 1875
int virQEMUCapsAddCPUDefinition(virQEMUCapsPtr qemuCaps,
                                const char *name)
1876
{
1877 1878 1879
    char *tmp;

    if (VIR_STRDUP(tmp, name) < 0)
1880
        return -1;
1881
    if (VIR_EXPAND_N(qemuCaps->cpuDefinitions, qemuCaps->ncpuDefinitions, 1) < 0) {
1882 1883 1884
        VIR_FREE(tmp);
        return -1;
    }
1885
    qemuCaps->cpuDefinitions[qemuCaps->ncpuDefinitions-1] = tmp;
1886 1887 1888 1889
    return 0;
}


1890 1891
size_t virQEMUCapsGetCPUDefinitions(virQEMUCapsPtr qemuCaps,
                                    char ***names)
1892
{
1893
    if (names)
1894 1895
        *names = qemuCaps->cpuDefinitions;
    return qemuCaps->ncpuDefinitions;
1896 1897 1898
}


1899 1900
size_t virQEMUCapsGetMachineTypes(virQEMUCapsPtr qemuCaps,
                                  char ***names)
1901
{
1902
    if (names)
1903 1904
        *names = qemuCaps->machineTypes;
    return qemuCaps->nmachineTypes;
1905 1906
}

1907 1908 1909
int virQEMUCapsGetMachineTypesCaps(virQEMUCapsPtr qemuCaps,
                                   size_t *nmachines,
                                   virCapsGuestMachinePtr **machines)
1910 1911 1912 1913 1914
{
    size_t i;

    *nmachines = 0;
    *machines = NULL;
1915
    if (VIR_ALLOC_N(*machines, qemuCaps->nmachineTypes) < 0)
1916
        goto error;
1917
    *nmachines = qemuCaps->nmachineTypes;
1918

1919
    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
1920 1921
        virCapsGuestMachinePtr mach;
        if (VIR_ALLOC(mach) < 0)
1922
            goto error;
1923
        if (qemuCaps->machineAliases[i]) {
1924 1925 1926
            if (VIR_STRDUP(mach->name, qemuCaps->machineAliases[i]) < 0 ||
                VIR_STRDUP(mach->canonical, qemuCaps->machineTypes[i]) < 0)
                goto error;
1927
        } else {
1928 1929
            if (VIR_STRDUP(mach->name, qemuCaps->machineTypes[i]) < 0)
                goto error;
1930
        }
1931
        mach->maxCpus = qemuCaps->machineMaxCpus[i];
1932 1933 1934 1935 1936
        (*machines)[i] = mach;
    }

    return 0;

1937
error:
1938 1939 1940 1941 1942 1943 1944 1945
    virCapabilitiesFreeMachines(*machines, *nmachines);
    *nmachines = 0;
    *machines = NULL;
    return -1;
}



1946

1947 1948
const char *virQEMUCapsGetCanonicalMachine(virQEMUCapsPtr qemuCaps,
                                           const char *name)
1949 1950 1951
{
    size_t i;

1952 1953 1954
    if (!name)
        return NULL;

1955
    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
1956
        if (!qemuCaps->machineAliases[i])
1957
            continue;
1958 1959
        if (STREQ(qemuCaps->machineAliases[i], name))
            return qemuCaps->machineTypes[i];
1960 1961 1962 1963
    }

    return name;
}
1964 1965


1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
int virQEMUCapsGetMachineMaxCpus(virQEMUCapsPtr qemuCaps,
                                 const char *name)
{
    size_t i;

    if (!name)
        return 0;

    for (i = 0; i < qemuCaps->nmachineTypes; i++) {
        if (!qemuCaps->machineMaxCpus[i])
            continue;
        if (STREQ(qemuCaps->machineTypes[i], name))
            return qemuCaps->machineMaxCpus[i];
    }

    return 0;
}


1985
static int
1986 1987
virQEMUCapsProbeQMPCommands(virQEMUCapsPtr qemuCaps,
                            qemuMonitorPtr mon)
1988 1989 1990 1991 1992 1993 1994
{
    char **commands = NULL;
    int ncommands;

    if ((ncommands = qemuMonitorGetCommands(mon, &commands)) < 0)
        return -1;

1995 1996 1997 1998 1999
    virQEMUCapsProcessStringFlags(qemuCaps,
                                  ARRAY_CARDINALITY(virQEMUCapsCommands),
                                  virQEMUCapsCommands,
                                  ncommands, commands);
    virQEMUCapsFreeStringList(ncommands, commands);
2000

2001 2002 2003 2004
    /* QMP add-fd was introduced in 1.2, but did not support
     * management control of set numbering, and did not have a
     * counterpart -add-fd command line option.  We require the
     * add-fd features from 1.3 or later.  */
2005
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_ADD_FD)) {
2006 2007 2008 2009 2010 2011 2012
        int fd = open("/dev/null", O_RDONLY);
        if (fd < 0) {
            virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                           _("unable to probe for add-fd"));
            return -1;
        }
        if (qemuMonitorAddFd(mon, 0, fd, "/dev/null") < 0)
2013
            virQEMUCapsClear(qemuCaps, QEMU_CAPS_ADD_FD);
2014 2015 2016
        VIR_FORCE_CLOSE(fd);
    }

2017 2018 2019 2020 2021
    return 0;
}


static int
2022 2023
virQEMUCapsProbeQMPEvents(virQEMUCapsPtr qemuCaps,
                          qemuMonitorPtr mon)
2024 2025 2026 2027 2028 2029 2030
{
    char **events = NULL;
    int nevents;

    if ((nevents = qemuMonitorGetEvents(mon, &events)) < 0)
        return -1;

2031 2032 2033 2034 2035
    virQEMUCapsProcessStringFlags(qemuCaps,
                                  ARRAY_CARDINALITY(virQEMUCapsEvents),
                                  virQEMUCapsEvents,
                                  nevents, events);
    virQEMUCapsFreeStringList(nevents, events);
2036 2037 2038 2039 2040

    return 0;
}


2041
static int
2042 2043
virQEMUCapsProbeQMPObjects(virQEMUCapsPtr qemuCaps,
                           qemuMonitorPtr mon)
2044 2045 2046 2047 2048 2049 2050
{
    int nvalues;
    char **values;
    size_t i;

    if ((nvalues = qemuMonitorGetObjectTypes(mon, &values)) < 0)
        return -1;
2051 2052 2053 2054 2055 2056
    virQEMUCapsProcessStringFlags(qemuCaps,
                                  ARRAY_CARDINALITY(virQEMUCapsObjectTypes),
                                  virQEMUCapsObjectTypes,
                                  nvalues, values);
    virQEMUCapsFreeStringList(nvalues, values);

2057
    for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsObjectProps); i++) {
2058
        const char *type = virQEMUCapsObjectProps[i].type;
2059 2060 2061 2062
        if ((nvalues = qemuMonitorGetObjectProps(mon,
                                                 type,
                                                 &values)) < 0)
            return -1;
2063 2064 2065 2066 2067
        virQEMUCapsProcessStringFlags(qemuCaps,
                                      virQEMUCapsObjectProps[i].nprops,
                                      virQEMUCapsObjectProps[i].props,
                                      nvalues, values);
        virQEMUCapsFreeStringList(nvalues, values);
2068 2069 2070
    }

    /* Prefer -chardev spicevmc (detected earlier) over -device spicevmc */
2071 2072
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEVMC))
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_DEVICE_SPICEVMC);
2073
    /* If qemu supports newer -device qxl it supports -vga qxl as well */
2074 2075
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE_QXL))
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA_QXL);
2076 2077 2078 2079 2080 2081

    return 0;
}


static int
2082 2083
virQEMUCapsProbeQMPMachineTypes(virQEMUCapsPtr qemuCaps,
                                qemuMonitorPtr mon)
2084 2085 2086 2087 2088
{
    qemuMonitorMachineInfoPtr *machines = NULL;
    int nmachines = 0;
    int ret = -1;
    size_t i;
2089
    size_t defIdx = 0;
2090 2091 2092 2093

    if ((nmachines = qemuMonitorGetMachines(mon, &machines)) < 0)
        goto cleanup;

2094
    if (VIR_ALLOC_N(qemuCaps->machineTypes, nmachines) < 0)
2095
        goto cleanup;
2096
    if (VIR_ALLOC_N(qemuCaps->machineAliases, nmachines) < 0)
2097
        goto cleanup;
2098
    if (VIR_ALLOC_N(qemuCaps->machineMaxCpus, nmachines) < 0)
2099
        goto cleanup;
2100

2101
    for (i = 0; i < nmachines; i++) {
2102 2103
        if (VIR_STRDUP(qemuCaps->machineAliases[i], machines[i]->alias) < 0 ||
            VIR_STRDUP(qemuCaps->machineTypes[i], machines[i]->name) < 0)
2104
            goto cleanup;
2105 2106
        if (machines[i]->isDefault)
            defIdx = i;
2107
        qemuCaps->machineMaxCpus[i] = machines[i]->maxCpus;
2108
    }
2109
    qemuCaps->nmachineTypes = nmachines;
2110 2111

    if (defIdx)
2112
        virQEMUCapsSetDefaultMachine(qemuCaps, defIdx);
2113 2114 2115 2116

    ret = 0;

cleanup:
2117
    for (i = 0; i < nmachines; i++)
2118 2119 2120 2121 2122 2123 2124
        qemuMonitorMachineInfoFree(machines[i]);
    VIR_FREE(machines);
    return ret;
}


static int
2125 2126
virQEMUCapsProbeQMPCPUDefinitions(virQEMUCapsPtr qemuCaps,
                                  qemuMonitorPtr mon)
2127 2128 2129 2130 2131 2132 2133
{
    int ncpuDefinitions;
    char **cpuDefinitions;

    if ((ncpuDefinitions = qemuMonitorGetCPUDefinitions(mon, &cpuDefinitions)) < 0)
        return -1;

2134 2135
    qemuCaps->ncpuDefinitions = ncpuDefinitions;
    qemuCaps->cpuDefinitions = cpuDefinitions;
2136 2137 2138 2139

    return 0;
}

2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162
struct tpmTypeToCaps {
    int type;
    enum virQEMUCapsFlags caps;
};

static const struct tpmTypeToCaps virQEMUCapsTPMTypesToCaps[] = {
    {
        .type = VIR_DOMAIN_TPM_TYPE_PASSTHROUGH,
        .caps = QEMU_CAPS_DEVICE_TPM_PASSTHROUGH,
    },
};

const struct tpmTypeToCaps virQEMUCapsTPMModelsToCaps[] = {
    {
        .type = VIR_DOMAIN_TPM_MODEL_TIS,
        .caps = QEMU_CAPS_DEVICE_TPM_TIS,
    },
};

static int
virQEMUCapsProbeQMPTPM(virQEMUCapsPtr qemuCaps,
                       qemuMonitorPtr mon)
{
2163 2164
    int nentries;
    size_t i;
2165
    char **entries = NULL;
S
Stefan Berger 已提交
2166

2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196
    if ((nentries = qemuMonitorGetTPMModels(mon, &entries)) < 0)
        return -1;

    if (nentries > 0) {
        for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsTPMModelsToCaps); i++) {
            const char *needle = virDomainTPMModelTypeToString(
                virQEMUCapsTPMModelsToCaps[i].type);
            if (virStringArrayHasString(entries, needle))
                virQEMUCapsSet(qemuCaps,
                               virQEMUCapsTPMModelsToCaps[i].caps);
        }
    }
    virStringFreeList(entries);

    if ((nentries = qemuMonitorGetTPMTypes(mon, &entries)) < 0)
        return -1;

    if (nentries > 0) {
        for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsTPMTypesToCaps); i++) {
            const char *needle = virDomainTPMBackendTypeToString(
                virQEMUCapsTPMTypesToCaps[i].type);
            if (virStringArrayHasString(entries, needle))
                virQEMUCapsSet(qemuCaps, virQEMUCapsTPMTypesToCaps[i].caps);
        }
    }
    virStringFreeList(entries);

    return 0;
}

2197

2198
static int
2199 2200
virQEMUCapsProbeQMPKVMState(virQEMUCapsPtr qemuCaps,
                            qemuMonitorPtr mon)
2201 2202 2203 2204
{
    bool enabled = false;
    bool present = false;

2205
    if (!virQEMUCapsGet(qemuCaps, QEMU_CAPS_KVM))
2206 2207 2208 2209 2210 2211
        return 0;

    if (qemuMonitorGetKVMState(mon, &enabled, &present) < 0)
        return -1;

    /* The QEMU_CAPS_KVM flag was initially set according to the QEMU
2212 2213 2214 2215 2216 2217 2218 2219
     * reporting the recognition of 'query-kvm' QMP command. That merely
     * indicates existance of the command though, not whether KVM support
     * is actually available, nor whether it is enabled by default.
     *
     * If it is not present we need to clear the flag, and if it is
     * not enabled by default we need to change the flag.
     */
    if (!present) {
2220
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_KVM);
2221
    } else if (!enabled) {
2222 2223
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_KVM);
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_ENABLE_KVM);
2224 2225 2226 2227 2228
    }

    return 0;
}

2229 2230 2231 2232 2233 2234 2235 2236
struct virQEMUCapsCommandLineProps {
    const char *option;
    const char *param;
    int flag;
};

static struct virQEMUCapsCommandLineProps virQEMUCapsCommandLine[] = {
    { "machine", "mem-merge", QEMU_CAPS_MEM_MERGE },
O
Osier Yang 已提交
2237
    { "drive", "discard", QEMU_CAPS_DRIVE_DISCARD },
2238
    { "realtime", "mlock", QEMU_CAPS_MLOCK },
2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264
};

static int
virQEMUCapsProbeQMPCommandLine(virQEMUCapsPtr qemuCaps,
                               qemuMonitorPtr mon)
{
    int nvalues;
    char **values;
    size_t i, j;

    for (i = 0; i < ARRAY_CARDINALITY(virQEMUCapsCommandLine); i++) {
        if ((nvalues = qemuMonitorGetCommandLineOptionParameters(mon,
                                                                 virQEMUCapsCommandLine[i].option,
                                                                 &values)) < 0)
            return -1;
        for (j = 0; j < nvalues; j++) {
            if (STREQ(virQEMUCapsCommandLine[i].param, values[j])) {
                virQEMUCapsSet(qemuCaps, virQEMUCapsCommandLine[i].flag);
                break;
            }
        }
        virStringFreeList(values);
    }

    return 0;
}
2265

2266 2267
int virQEMUCapsProbeQMP(virQEMUCapsPtr qemuCaps,
                        qemuMonitorPtr mon)
2268
{
2269
    VIR_DEBUG("qemuCaps=%p mon=%p", qemuCaps, mon);
2270

2271
    if (qemuCaps->usedQMP)
2272 2273
        return 0;

2274
    if (virQEMUCapsProbeQMPCommands(qemuCaps, mon) < 0)
2275 2276
        return -1;

2277
    if (virQEMUCapsProbeQMPEvents(qemuCaps, mon) < 0)
2278 2279 2280 2281 2282 2283
        return -1;

    return 0;
}


2284 2285
#define QEMU_SYSTEM_PREFIX "qemu-system-"

2286
static int
2287
virQEMUCapsInitHelp(virQEMUCapsPtr qemuCaps, uid_t runUid, gid_t runGid)
2288
{
2289
    virCommandPtr cmd = NULL;
2290
    bool is_kvm;
2291
    char *help = NULL;
2292 2293
    int ret = -1;
    const char *tmp;
2294

2295
    VIR_DEBUG("qemuCaps=%p", qemuCaps);
2296

2297
    tmp = strstr(qemuCaps->binary, QEMU_SYSTEM_PREFIX);
2298 2299
    if (tmp) {
        tmp += strlen(QEMU_SYSTEM_PREFIX);
2300

2301
        qemuCaps->arch = virQEMUCapsArchFromString(tmp);
2302
    } else {
2303
        qemuCaps->arch = virArchFromHost();
2304 2305
    }

2306
    cmd = virQEMUCapsProbeCommand(qemuCaps->binary, NULL, runUid, runGid);
2307 2308 2309 2310
    virCommandAddArgList(cmd, "-help", NULL);
    virCommandSetOutputBuffer(cmd, &help);

    if (virCommandRun(cmd, NULL) < 0)
2311
        goto cleanup;
2312

2313 2314 2315 2316 2317 2318
    if (virQEMUCapsParseHelpStr(qemuCaps->binary,
                                help, qemuCaps,
                                &qemuCaps->version,
                                &is_kvm,
                                &qemuCaps->kvmVersion,
                                false) < 0)
2319
        goto cleanup;
2320 2321

    /* Currently only x86_64 and i686 support PCI-multibus. */
2322 2323 2324
    if (qemuCaps->arch == VIR_ARCH_X86_64 ||
        qemuCaps->arch == VIR_ARCH_I686) {
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCI_MULTIBUS);
2325 2326 2327
    } else {
        /* -no-acpi is not supported on other archs
         * even if qemu reports it in -help */
2328
        virQEMUCapsClear(qemuCaps, QEMU_CAPS_NO_ACPI);
2329
    }
2330

2331
    /* virQEMUCapsExtractDeviceStr will only set additional caps if qemu
2332
     * understands the 0.13.0+ notion of "-device driver,".  */
2333
    if (virQEMUCapsGet(qemuCaps, QEMU_CAPS_DEVICE) &&
2334
        strstr(help, "-device driver,?") &&
2335 2336
        virQEMUCapsExtractDeviceStr(qemuCaps->binary,
                                    qemuCaps, runUid, runGid) < 0) {
2337
        goto cleanup;
2338
    }
2339

2340
    if (virQEMUCapsProbeCPUModels(qemuCaps, runUid, runGid) < 0)
2341
        goto cleanup;
2342

2343
    if (virQEMUCapsProbeMachineTypes(qemuCaps, runUid, runGid) < 0)
2344
        goto cleanup;
2345

2346
    ret = 0;
2347
cleanup:
2348
    virCommandFree(cmd);
2349
    VIR_FREE(help);
2350 2351 2352 2353
    return ret;
}


2354 2355
static void virQEMUCapsMonitorNotify(qemuMonitorPtr mon ATTRIBUTE_UNUSED,
                                     virDomainObjPtr vm ATTRIBUTE_UNUSED)
2356 2357 2358 2359
{
}

static qemuMonitorCallbacks callbacks = {
2360 2361
    .eofNotify = virQEMUCapsMonitorNotify,
    .errorNotify = virQEMUCapsMonitorNotify,
2362 2363 2364 2365 2366 2367 2368
};


/* Capabilities that we assume are always enabled
 * for QEMU >= 1.2.0
 */
static void
2369
virQEMUCapsInitQMPBasic(virQEMUCapsPtr qemuCaps)
2370
{
2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_COLON);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_REBOOT);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NAME);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_UUID);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNET_HDR);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_TCP);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_EXEC);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_V2);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_FORMAT);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_0_10);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MEM_PATH);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_SERIAL);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_UNIX);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MONITOR_JSON);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_BALLOON);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_SDL);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_SMP_TOPOLOGY);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_RTC);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VHOST_NET);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_HPET);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NODEFCONFIG);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_BOOT_MENU);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NAME_PROCESS);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_READONLY);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_SMBIOS_TYPE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VGA_NONE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MIGRATE_QEMU_FD);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_AIO);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_CHARDEV_SPICEVMC);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE_QXL_VGA);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_DIRECTSYNC);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_SHUTDOWN);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_CACHE_UNSAFE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV_READONLY);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VIRTIO_BLK_SG_IO);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_COPY_ON_READ);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_CPU_HOST);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_FSDEV_WRITEOUT);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DRIVE_IOTUNE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_WAKEUP);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_USER_CONFIG);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_NETDEV_BRIDGE);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_SECCOMP_SANDBOX);
O
Olivia Yin 已提交
2420
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DTB);
J
Ján Tomko 已提交
2421
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_IPV6_MIGRATION);
2422 2423
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_MACHINE_OPT);
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DUMP_GUEST_CORE);
2424
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_SHARE_POLICY);
2425
    virQEMUCapsSet(qemuCaps, QEMU_CAPS_DEVICE_VIDEO_PRIMARY);
2426 2427
}

2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463
/* Capabilities that are architecture depending
 * initialized for QEMU.
 */
static int
virQEMUCapsInitArchQMPBasic(virQEMUCapsPtr qemuCaps,
                            qemuMonitorPtr mon)
{
    char *archstr = NULL;
    int ret = -1;

    if (!(archstr = qemuMonitorGetTargetArch(mon)))
        return -1;

    if ((qemuCaps->arch = virQEMUCapsArchFromString(archstr)) == VIR_ARCH_NONE) {
        virReportError(VIR_ERR_INTERNAL_ERROR,
                       _("Unknown QEMU arch %s"), archstr);
        goto cleanup;
    }

    /*
     * Currently only x86_64 and i686 support PCI-multibus,
     * -no-acpi and -no-kvm-pit-reinjection.
     */
    if (qemuCaps->arch == VIR_ARCH_X86_64 ||
        qemuCaps->arch == VIR_ARCH_I686) {
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_PCI_MULTIBUS);
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_ACPI);
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_NO_KVM_PIT);
    }

    ret = 0;

cleanup:
    VIR_FREE(archstr);
    return ret;
}
2464 2465

static int
2466 2467 2468 2469
virQEMUCapsInitQMP(virQEMUCapsPtr qemuCaps,
                   const char *libDir,
                   uid_t runUid,
                   gid_t runGid)
2470 2471 2472 2473 2474
{
    int ret = -1;
    virCommandPtr cmd = NULL;
    qemuMonitorPtr mon = NULL;
    int major, minor, micro;
2475
    char *package = NULL;
2476 2477 2478 2479
    int status = 0;
    virDomainChrSourceDef config;
    char *monarg = NULL;
    char *monpath = NULL;
2480
    char *pidfile = NULL;
2481 2482
    pid_t pid = 0;
    virDomainObj vm;
2483

2484 2485 2486
    /* the ".sock" sufix is important to avoid a possible clash with a qemu
     * domain called "capabilities"
     */
2487
    if (virAsprintf(&monpath, "%s/%s", libDir, "capabilities.monitor.sock") < 0)
2488
        goto cleanup;
2489
    if (virAsprintf(&monarg, "unix:%s,server,nowait", monpath) < 0)
2490 2491
        goto cleanup;

2492 2493
    /* ".pidfile" suffix is used rather than ".pid" to avoid a possible clash
     * with a qemu domain called "capabilities"
2494 2495 2496
     * Normally we'd use runDir for pid files, but because we're using
     * -daemonize we need QEMU to be allowed to create them, rather
     * than libvirtd. So we're using libDir which QEMU can write to
2497
     */
2498
    if (virAsprintf(&pidfile, "%s/%s", libDir, "capabilities.pidfile") < 0)
2499 2500
        goto cleanup;

2501 2502 2503 2504 2505
    memset(&config, 0, sizeof(config));
    config.type = VIR_DOMAIN_CHR_TYPE_UNIX;
    config.data.nix.path = monpath;
    config.data.nix.listen = false;

2506
    VIR_DEBUG("Try to get caps via QMP qemuCaps=%p", qemuCaps);
2507

2508 2509 2510 2511 2512 2513 2514
    /*
     * We explicitly need to use -daemonize here, rather than
     * virCommandDaemonize, because we need to synchronize
     * with QEMU creating its monitor socket API. Using
     * daemonize guarantees control won't return to libvirt
     * until the socket is present.
     */
2515
    cmd = virCommandNewArgList(qemuCaps->binary,
2516 2517 2518 2519 2520 2521
                               "-S",
                               "-no-user-config",
                               "-nodefaults",
                               "-nographic",
                               "-M", "none",
                               "-qmp", monarg,
2522 2523
                               "-pidfile", pidfile,
                               "-daemonize",
2524 2525 2526
                               NULL);
    virCommandAddEnvPassCommon(cmd);
    virCommandClearCaps(cmd);
2527 2528
    virCommandSetGID(cmd, runGid);
    virCommandSetUID(cmd, runUid);
2529 2530 2531 2532 2533 2534

    if (virCommandRun(cmd, &status) < 0)
        goto cleanup;

    if (status != 0) {
        ret = 0;
2535
        VIR_DEBUG("QEMU %s exited with status %d", qemuCaps->binary, status);
2536 2537 2538
        goto cleanup;
    }

2539 2540 2541 2542 2543 2544 2545 2546 2547 2548
    if (virPidFileReadPath(pidfile, &pid) < 0) {
        VIR_DEBUG("Failed to read pidfile %s", pidfile);
        ret = 0;
        goto cleanup;
    }

    memset(&vm, 0, sizeof(vm));
    vm.pid = pid;

    if (!(mon = qemuMonitorOpen(&vm, &config, true, &callbacks))) {
2549
        ret = 0;
2550
        goto cleanup;
2551
    }
2552

2553
    virObjectLock(mon);
2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575

    if (qemuMonitorSetCapabilities(mon) < 0) {
        virErrorPtr err = virGetLastError();
        VIR_DEBUG("Failed to set monitor capabilities %s",
                  err ? err->message : "<unknown problem>");
        ret = 0;
        goto cleanup;
    }

    if (qemuMonitorGetVersion(mon,
                              &major, &minor, &micro,
                              &package) < 0) {
        virErrorPtr err = virGetLastError();
        VIR_DEBUG("Failed to query monitor version %s",
                  err ? err->message : "<unknown problem>");
        ret = 0;
        goto cleanup;
    }

    VIR_DEBUG("Got version %d.%d.%d (%s)",
              major, minor, micro, NULLSTR(package));

2576
    if (major < 1 || (major == 1 && minor < 2)) {
2577 2578 2579 2580 2581
        VIR_DEBUG("Not new enough for QMP capabilities detection");
        ret = 0;
        goto cleanup;
    }

2582 2583
    qemuCaps->version = major * 1000000 + minor * 1000 + micro;
    qemuCaps->usedQMP = true;
2584

2585
    virQEMUCapsInitQMPBasic(qemuCaps);
2586

2587 2588 2589
    if (virQEMUCapsInitArchQMPBasic(qemuCaps, mon) < 0)
        goto cleanup;

L
Li Zhang 已提交
2590 2591 2592 2593
    /* USB option is supported v1.3.0 onwards */
    if (qemuCaps->version >= 1003000)
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_MACHINE_USB_OPT);

2594 2595 2596 2597
    /* WebSockets were introduced between 1.3.0 and 1.3.1 */
    if (qemuCaps->version >= 1003001)
        virQEMUCapsSet(qemuCaps, QEMU_CAPS_VNC_WEBSOCKET);

2598
    if (virQEMUCapsProbeQMPCommands(qemuCaps, mon) < 0)
2599
        goto cleanup;
2600
    if (virQEMUCapsProbeQMPEvents(qemuCaps, mon) < 0)
2601
        goto cleanup;
2602
    if (virQEMUCapsProbeQMPObjects(qemuCaps, mon) < 0)
2603
        goto cleanup;
2604
    if (virQEMUCapsProbeQMPMachineTypes(qemuCaps, mon) < 0)
2605
        goto cleanup;
2606
    if (virQEMUCapsProbeQMPCPUDefinitions(qemuCaps, mon) < 0)
2607
        goto cleanup;
2608
    if (virQEMUCapsProbeQMPKVMState(qemuCaps, mon) < 0)
2609
        goto cleanup;
2610 2611
    if (virQEMUCapsProbeQMPTPM(qemuCaps, mon) < 0)
        goto cleanup;
2612 2613
    if (virQEMUCapsProbeQMPCommandLine(qemuCaps, mon) < 0)
        goto cleanup;
2614 2615 2616 2617 2618

    ret = 0;

cleanup:
    if (mon)
2619
        virObjectUnlock(mon);
2620 2621
    qemuMonitorClose(mon);
    virCommandAbort(cmd);
2622
    virCommandFree(cmd);
2623 2624
    VIR_FREE(monarg);
    VIR_FREE(monpath);
2625
    VIR_FREE(package);
2626

2627
    if (pid != 0) {
2628 2629
        char ebuf[1024];

2630 2631 2632 2633 2634 2635 2636
        VIR_DEBUG("Killing QMP caps process %lld", (long long) pid);
        if (virProcessKill(pid, SIGKILL) < 0 && errno != ESRCH)
            VIR_ERROR(_("Failed to kill process %lld: %s"),
                      (long long) pid,
                      virStrerror(errno, ebuf, sizeof(ebuf)));
    }
    if (pidfile) {
2637 2638 2639
        unlink(pidfile);
        VIR_FREE(pidfile);
    }
2640 2641 2642 2643
    return ret;
}


2644 2645 2646 2647
virQEMUCapsPtr virQEMUCapsNewForBinary(const char *binary,
                                       const char *libDir,
                                       uid_t runUid,
                                       gid_t runGid)
2648
{
2649
    virQEMUCapsPtr qemuCaps = virQEMUCapsNew();
2650 2651 2652
    struct stat sb;
    int rv;

2653 2654
    if (VIR_STRDUP(qemuCaps->binary, binary) < 0)
        goto error;
2655 2656 2657 2658 2659 2660 2661 2662

    /* We would also want to check faccessat if we cared about ACLs,
     * but we don't.  */
    if (stat(binary, &sb) < 0) {
        virReportSystemError(errno, _("Cannot check QEMU binary %s"),
                             binary);
        goto error;
    }
2663
    qemuCaps->mtime = sb.st_mtime;
2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674

    /* Make sure the binary we are about to try exec'ing exists.
     * Technically we could catch the exec() failure, but that's
     * in a sub-process so it's hard to feed back a useful error.
     */
    if (!virFileIsExecutable(binary)) {
        virReportSystemError(errno, _("QEMU binary %s is not executable"),
                             binary);
        goto error;
    }

2675
    if ((rv = virQEMUCapsInitQMP(qemuCaps, libDir, runUid, runGid)) < 0)
2676 2677
        goto error;

2678 2679
    if (!qemuCaps->usedQMP &&
        virQEMUCapsInitHelp(qemuCaps, runUid, runGid) < 0)
2680 2681
        goto error;

2682
    return qemuCaps;
2683 2684

error:
2685 2686
    virObjectUnref(qemuCaps);
    qemuCaps = NULL;
2687
    return NULL;
2688 2689 2690
}


2691
bool virQEMUCapsIsValid(virQEMUCapsPtr qemuCaps)
2692 2693 2694
{
    struct stat sb;

2695
    if (!qemuCaps->binary)
2696 2697
        return true;

2698
    if (stat(qemuCaps->binary, &sb) < 0)
2699 2700
        return false;

2701
    return sb.st_mtime == qemuCaps->mtime;
2702
}
2703 2704 2705


static void
2706
virQEMUCapsHashDataFree(void *payload, const void *key ATTRIBUTE_UNUSED)
2707 2708 2709 2710 2711
{
    virObjectUnref(payload);
}


2712 2713 2714 2715
virQEMUCapsCachePtr
virQEMUCapsCacheNew(const char *libDir,
                    uid_t runUid,
                    gid_t runGid)
2716
{
2717
    virQEMUCapsCachePtr cache;
2718

2719
    if (VIR_ALLOC(cache) < 0)
2720 2721 2722 2723 2724 2725 2726 2727 2728
        return NULL;

    if (virMutexInit(&cache->lock) < 0) {
        virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
                       _("Unable to initialize mutex"));
        VIR_FREE(cache);
        return NULL;
    }

2729
    if (!(cache->binaries = virHashCreate(10, virQEMUCapsHashDataFree)))
2730
        goto error;
2731
    if (VIR_STRDUP(cache->libDir, libDir) < 0)
2732
        goto error;
2733

2734 2735 2736
    cache->runUid = runUid;
    cache->runGid = runGid;

2737 2738 2739
    return cache;

error:
2740
    virQEMUCapsCacheFree(cache);
2741 2742 2743 2744
    return NULL;
}


2745 2746
virQEMUCapsPtr
virQEMUCapsCacheLookup(virQEMUCapsCachePtr cache, const char *binary)
2747
{
2748
    virQEMUCapsPtr ret = NULL;
2749 2750 2751
    virMutexLock(&cache->lock);
    ret = virHashLookup(cache->binaries, binary);
    if (ret &&
2752
        !virQEMUCapsIsValid(ret)) {
2753 2754 2755 2756 2757 2758 2759 2760
        VIR_DEBUG("Cached capabilities %p no longer valid for %s",
                  ret, binary);
        virHashRemoveEntry(cache->binaries, binary);
        ret = NULL;
    }
    if (!ret) {
        VIR_DEBUG("Creating capabilities for %s",
                  binary);
2761 2762
        ret = virQEMUCapsNewForBinary(binary, cache->libDir,
                                      cache->runUid, cache->runGid);
2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778
        if (ret) {
            VIR_DEBUG("Caching capabilities %p for %s",
                      ret, binary);
            if (virHashAddEntry(cache->binaries, binary, ret) < 0) {
                virObjectUnref(ret);
                ret = NULL;
            }
        }
    }
    VIR_DEBUG("Returning caps %p for %s", ret, binary);
    virObjectRef(ret);
    virMutexUnlock(&cache->lock);
    return ret;
}


2779 2780
virQEMUCapsPtr
virQEMUCapsCacheLookupCopy(virQEMUCapsCachePtr cache, const char *binary)
2781
{
2782 2783
    virQEMUCapsPtr qemuCaps = virQEMUCapsCacheLookup(cache, binary);
    virQEMUCapsPtr ret;
2784

2785
    if (!qemuCaps)
2786 2787
        return NULL;

2788 2789
    ret = virQEMUCapsNewCopy(qemuCaps);
    virObjectUnref(qemuCaps);
2790 2791 2792 2793 2794
    return ret;
}


void
2795
virQEMUCapsCacheFree(virQEMUCapsCachePtr cache)
2796 2797 2798 2799
{
    if (!cache)
        return;

2800
    VIR_FREE(cache->libDir);
2801 2802 2803 2804
    virHashFree(cache->binaries);
    virMutexDestroy(&cache->lock);
    VIR_FREE(cache);
}
2805 2806

bool
2807
virQEMUCapsUsedQMP(virQEMUCapsPtr qemuCaps)
2808
{
2809
    return qemuCaps->usedQMP;
2810
}